SITALWeek #395

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: a close look at the Hollywood writers' strike and its broader ramifications for AI displacement; integrating reward and punishment into AI feedback loops could advance models far more rapidly; the value of surprise; speaking of surprises, apparently mind reading is becoming a reality; and, much more below.

Stuff about Innovation and Technology
AI Goes to Hollywood
Hollywood’s Writers Guild chose to go on strike last week after failed negotiations with the major studios. Among the complaints are concerns regarding the way streaming has negatively impacted residuals and reduced the size/duration of writers’ rooms, as well as the threat posed by AI’s expanding role. AI’s impact on content production has been a frequent topic in SITALWeek, as I see the movie industry as a hallmark for how AI might impact other industries both inside and outside of the creative arts. Whether it’s de-aging, voice-overs, AI scripting, virtual world sets, or, ultimately, actors selling rights to their entire persona (as in the 2013 sci-fi movie The Congress), it’s an industry ripe for disruption. Hollywood has always been an early tech adopter, and the industry long benefited from steady growth in viewing as screens proliferated and mobile devices expanded consumers’ screen time. However, recent developments have left us awash in a sea of infinite content with offerings that far outnumber the professional gems produced by Hollywood and other studios. Indeed, I think it’s fair to say that scripted content is becoming a niche industry that is at risk of shrinking its share of viewing even further with the current work stoppage. Even if there is a prolonged cessation (which seems likely given the negotiation gulf between the writers and the studios), we consumers have nothing to worry about because, as Matt Belloni from Puck joked recently: “the algorithms will take care of us”. There’s probably a lot of truth in that quip given the massive amount of content across YouTube, TikTok, podcasts, video games, unscripted shows, streaming services’ library content, and already scripted shows in production. Just last week, The Information reported that 45% of US YouTube viewing is now taking place on TV screens, up from less than 30% in 2020. 

Writers' desire to draw up some rules for how AI will be used by Hollywood makes sense, but it would be a mistake to believe it won’t have a large – and potentially negative – impact on the demand for their services. This could be bad timing for the strike, as The Hollywood Reporter points out, because AI can cross the picket line (which raises an awkward futuristic thought experiment: should AI be allowed to unionize?). I’d like to think that uniquely human sparks of creativity and connection (between writers, directors, actors, editors, composers, etc.) are responsible for the magic of great movies and TV shows, but I’m also open to the possibility that advanced AI will be able to match (or maybe even surpass) the quality of human creativity with time. Further, the future of storytelling might end up being more interactive and personal, requiring not so much a human writer, but a God-like AI engine that creates virtual worlds, stories, or even movie adaptations of favorite books customized for each viewer on demand. In a recent interview, Joe Russo, the co-director of Avengers: Endgame, and Donald Mustard, the Chief Creative Officer at Epic Games (maker of the Unreal Engine that is also used for virtual stages to shoot movies and TV shows), discussed this future of AI and storytelling: 
Russo: So potentially, what you could do with [AI] is obviously use it to engineer storytelling and change storytelling. So you have a constantly evolving story, either in a game or in a movie, or a TV show. You could walk into your house and say to the AI on your streaming platform: “Hey, I want a movie starring my photoreal avatar and Marilyn Monroe's photoreal avatar. I want it to be a rom-com because I've had a rough day,” and it renders a very competent story with dialogue that mimics your voice. It mimics your voice, and suddenly now you have a rom-com starring you that's 90 minutes long. So you can curate your story specifically to you.
That's one thing that it can do, but it can also, on a communal level, populate the world of the game, have intelligence behind character choice, you know, the computer-run characters in the game that can make decisions learn your play style, make it a little harder for you, make it a little easier for you, curate the story...How quickly we get there, I don't know, but that's where it's going.
Mustard: ...we're really not that far off from where some of the real-time engines like Unreal Engine...[are] very close to where you could be almost perfect photorealistic, real-time rendering...Or, on the fly, you could be like, “Yep, I wanna star as myself in a movie,” and you could watch it.

The writers’ strike is just one example of the discord we will see as workers and employers adapt to rapidly evolving technology. Every profession needs to be ready to deal with the impacts of AI on jobs and business models by identifying places where people can still add value or work alongside AI. A good starting point with any disruptive force is to ask as many questions as possible in order to plumb the limits of possible scenarios and outcomes. An important part of that process is having an open mind and holding your previous beliefs very loosely, a topic I explored in more detail in More Q, Less A.

AI Pile Drivers, Back Office, and TSA
A trio of stories got me thinking about how quickly some jobs will be taken over by AI and automation. Built Robotics RPD 35 robot is a fully autonomous piling machine that can set the stage for large-scale solar installations by putting piles into the ground 3-5x faster with more accuracy than human operators and crew (the bot can carry 200 15-foot beams and drive one eight feet into the ground every 78 seconds). Also in the news, the IBM CEO indicated they are pausing 7,800 hires that they think can be replaced by AI. Lastly, the TSA is rolling out automated facial recognition tools so travelers can self-scan IDs for authentication without interacting with an agent. I am reminded of the significant transfer of Western jobs to India-based IT and business process outsourcing (BPO) firms a couple of decades ago. That was a painful, but generally slow enough process that our job markets adapted despite some permanent job losses. Many of those jobs that were offshored are in customer service and back office, two areas that appear to be rapidly set for disruption from AI.

Miscellaneous Stuff
Surprise and Reward
Most people believe that their thoughts and actions are reactionary to the surrounding environment, but, in reality, our brain predicts what will happen second to second and then changes our behavior based on whether or not the predictions are correct. We are rewarded for correct predictions by a sense of order and balance, but we also need to push our limits a bit (and risk being wrong) in order to gain more experience on which to base future predictions. Recently, cognitive philosopher Andy Clark and cosmologist Sean Carroll had an insightful, high-level discussion on Carroll’s podcast of many of the reasons for (and consequences of) this prediction-based modus operandi and how this lens helps us understand human behavior. One part of their discussion that I connected with was the importance of surprise. Here is an excerpt of the conversation:
Sean Carroll: So couldn't we just say that there are two things going on? We want to minimize prediction error, but we also want to survive, so there's a constraint, we want to survive, and under that constraint, it's actually useful to go out and be surprised sometimes so we can update our predictive model...I don't know how mathematically that will work out, but it does seem a little bit intuitive to me.
Andy Clark: ...it looks as if very often, the correct move for a prediction-driven system is to temporarily increase its own uncertainty so as to do a better job over the long time scale of minimizing prediction errors, and that looks like the value of surprise, actually...I think we artificially curate environments in which we can surprise ourselves. I think, actually, this is maybe what art and science is to some extent, at least, we're curating environments in which we can harvest the kind of surprises that improve our generative models, our understandings of the world in ways that enable us to be less surprised about certain things in the future.
There is an obvious connection between this predictive model of the brain and large language models, which themselves are a form of predictive autocomplete. As we learned from Wolfram, one of the interesting things about LLMs that make them appear more creative is they don’t always choose the most obvious next word, i.e., they have a built-in element of surprise (see You Auto-Complete Me for more). Clark suggests in the podcast that LLMs suffer from not having a proper reward feedback loop, i.e., they aren’t rewarded for providing useful answers and there are no consequences for bad behavior (lies, insults, etc.). It would therefore be interesting to add reward-based training and operation for LLMs. This is indeed a key idea advanced by Karl Friston, one of the original proponents of the prediction model of the brain (see also #271 and #272). I’ve also previously covered an Australian company working on neural nets comprised of mouse neurons that rewards the cell networks with predictable signals and punishes them with unpredictable ones (#370). I think we could achieve some very interesting human-like AI by combining embodied LLMs (having a physical form is key to accessing environmental inputs and having parameters for seeking balance) with reward/punishment-based reinforcement learning. I suspect it will become especially critical to reward robots for good behavior as LLMs and AI enter the physical world and interact with humans.

Programmed vs. Free-Thinking Bots
Speaking of human-robot interactions, Lex Fridman’s interview with Boston Dynamics CEO Robert Playter was interesting. I will say, however, that I tend to disagree with Playter’s focus, as he voices more interest in pre-programmed, rote robotic tasks than in embodying AI in autonomous form factors capable of learning on the fly. Boston Dynamics has put its AI efforts into a separate division that is now run by BD’s founder. I think the biggest near-term advancements are more likely to come from a deep integration of AI and robotics, where AI can learn by interacting with the physical world (see AI Awareness).

Mind Reading is Real!?
Scientists have successfully used an early version of GPT to read human thoughts and translate them into text. This incredible breakthrough was published in Nature Neuroscience. The technique is not a general-purpose mind reading machine due to the nature of the human brain and fMRI tools (see Whole-Brain Signaling for more on that topic). Rather, it needs to be tuned and trained on each person. But, once trained, it can effectively translate what you are thinking into a decipherable text output. Intriguingly, the output is not word for word, but rather captures the spirit of the thoughts. This too reminds me of how LLMs’ predictive autocomplete often chooses different words to communicate the same thing. Functional MRIs are sophisticated machines that are clearly not portable; but, I suspect that, with enough training and GPT/hardware advances, something more akin to a wearable device could eventually voice your thoughts by reading your brain activity. There are obvious dangers to such a technology, but this advance is truly mind blowing to me, and it seems to reveal a lot about just how easy it might be to decipher (and replicate in silico) the complexity of the human brain. 

Deep Water
I enjoyed the 2022 David Bowie doc Moonage Daydream released on HBO Max last week. That led me also to watch David Bowie: The Last Five Years (also currently available on HBO Max). That film contained a Bowie quote that feels appropriate in many ways right now, especially as it relates to the speed of AI disruption and the importance of surprise and randomness to finding the right questions to ask: “If you feel safe in the area you’re working in, you’re not working in the right area. Always go a little further into the water than you feel you’re capable of being in. Go a little bit out of your depth. And when you don’t feel that your feet are quite touching the bottom, you’re just about in the right place to do something exciting.”

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #394

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: chatbot vs. chatbot; understanding LLMs as a new way to interact with technology that gives users more agency; quantum sensors; ecommerce and digital ad giants enabled decades of innovation in new products and services, but as growth slows and focus shifts to cost cutting, who will create the next platforms for disruption?; Buster Poindexter; and, much more below.

Stuff about Innovation and Technology
AI Negotiators
Walmart’s AI chatbot can simultaneously converse with 2,000 suppliers to extract...er, negotiate, as Walmart puts it...improved terms. The bot learns on the fly how to maximize tradeoffs, like taking lower prices in order to get paid faster or vice versa, and it can apply what it learns to future interactions. Other companies, like shipping giant Maersk, also deploy AI negotiation tools, according to Supply Management. Amazon has been leveraging AI with mixed outcomes for years with their “hands off the wheel” strategy. The best counterdefense to AI-negotiating chatbots would probably be deployment of your own chatbot advocate. I currently have chatbots negotiating some of my recurring monthly subscription bills. Before long, most customer relations will be chatbots talking to chatbots – the hands-off-the-wheel future that will greatly increase volatility for us humans who will be experiencing the outcomes of these silent battles. One area that might see significant impact from chatbot-to-chatbot negotiations is complex legal contracts, as Fortune reports that many startups and law firms are already leveraging AI tech. 

LLMs as Social Collaboration Interface
This explanation of large language model AI from Jaron Lanier is compelling from a number of angles. Here is an excerpt from the essay in the New Yorker, which I recommend reading in full:
If the new tech isn’t true artificial intelligence, then what is it? In my view, the most accurate way to understand what we are building today is as an innovative form of social collaboration.
A program like OpenAI’s GPT-4, which can write sentences to order, is something like a version of Wikipedia that includes much more data, mashed together using statistics. Programs that create images to order are something like a version of online image search, but with a system for combining the pictures. In both cases, it’s people who have written the text and furnished the images. The new programs mash up work done by human minds. What’s innovative is that the mashup process has become guided and constrained, so that the results are usable and often striking. This is a significant achievement and worth celebrating—but it can be thought of as illuminating previously hidden concordances between human creations, rather than as the invention of a new mind.
As far as I can tell, my view flatters the technology. After all, what is civilization but social collaboration? Seeing A.I. as a way of working together, rather than as a technology for creating independent, intelligent beings, may make it less mysterious—less like HAL 9000 or Commander Data. But that’s good, because mystery only makes mismanagement more likely.
Further, here’s Lanier’s characterization of how digital platforms force us to conform to them, and how AI may flip that:
Many of the uses of A.I. that I like rest on advantages we gain when computers get less rigid. Digital stuff as we have known it has a brittle quality that forces people to conform to it, rather than assess it. We’ve all endured the agony of watching some poor soul at a doctor’s office struggle to do the expected thing on a front-desk screen. The face contorts; humanity is undermined. The need to conform to digital designs has created an ambient expectation of human subservience. A positive spin on A.I. is that it might spell the end of this torture, if we use it well.

Squeezed, Entangled Photons for Navigation
Researchers at the Universities of Arizona and Michigan have used distributed quantum sensing and squeezed light to improve the speed and precision of light-based accelerometers by 60% and 40%, respectively. These optomechanical sensors may prove invaluable in navigational systems where GPS performance is limited (e.g., underground, underwater, indoors, outer space). The advance uses quantum entanglement to split a laser beam into two, which can then be analyzed by two sensors as if it were the same beam (owing to its photons being entangled with each other) to improve precision (the more times you can measure the same thing, the more precise the measurement). Squeezed light reduces the quantum uncertainty in one parameter, like the photonic phase, at the expense of sensitivity for measuring other, extraneous variables to further enhance precision. While quantum computing remains an experimental pipe dream, quantum sensors and quantum communication networks will likely be commercialized sooner. 

They’re Cute…Until They’re Not
This IEEE article highlights three new robotic prototypes. I cannot get enough of these adorable little robots doing cute things like playing sports, even though, one day, they will probably hunt down every last member of humankind. It feels like an incredibly short distance between today’s experimental robots and Black Mirror’s graphic robotic dog attack in “Metalhead”. I am both looking forward to and dreading the day that these little bots are loaded up with AI. (For more on this topic see last week’s Soccerbots). In the meantime, the clairvoyant show Black Mirror is set to return to Netflix for season six this summer. Regarding the show’s four-year break, creator Charlie Brooker described the pandemic-weary world as not having the stomach for new episodes.

Passing the Innovation Baton
In aggregate, the three big pillars of the Internet – digital ads (search, social, etc.), ecommerce, and streaming video – have experienced a slowdown in revenue growth in recent quarters. While initially attributed to lapping the post-pandemic demand pull in, some elements of the slowdown also appear to be related to the maturity of the decades-old industries and user behavior. This malaise has been met with widespread cost cutting efforts at companies like Meta, Alphabet, and Amazon. Back in November, in a long post entitled End of Advertising-Funded R&D?, I reflected on how maturing s-curves in digital ad businesses might curtail future experimental funding in new technologies. Over the last few months, my thinking on this topic has expanded to encompass a more general, weaker backdrop for innovation fueled by the mega platforms. For two decades, the creation of the digital advertising, ecommerce, and mobile phone markets (especially by Google, Meta, Amazon, and Apple) was a primary engine of asymmetry and new business model formation across a host of products and services. AWS, created and funded by Amazon’s ecommerce business, became the platform for nearly every new tech startup in industries from media to fintech. Businesses like Netflix and Uber were built on cloud platforms like AWS (and smartphone apps running on Android and iOS). We might have eventually arrived at these new digitally-enabled businesses, but it’s possible the tens of billions of dollars needed to build out the infrastructure, subsidized by advertising and ecommerce, was an existential enabler. Likewise, without Azure (which was a response to the competitive threat of AWS and GCP/Google Suite) and Microsoft’s $billions, we likely wouldn’t already have OpenAI’s world-altering ChatGPT. It’s hard to separate out the causes of today’s slowdown for the mega platforms, but it could be a reflection that we have maxed out our potential screen time and have tapped much of the potential of online shopping. Perhaps growth will be a little harder to come by, or perhaps AI represents an entirely new mega platform to subsidize a vast new array of businesses we cannot even imagine today. I’m by no means ready to posthumously eulogize today’s digital giants. If anything, they remain the key enabling layer of the future of the economy, including AI. However, as these platforms experience a slowdown, or perhaps even a de-powerlawing, innovation is shifting from the unbounded, positive feedback loops of the purely digital world to the messier negative feedback loops of the physical world. As this happens, innovation will make inroads into the analog infrastructure layers of the economy. Digital innovation and analog transformation are like two gears spinning at different speeds, and friction inevitably arises at the interface. AI has the potential to greatly speed up the digital gear to a point where the analog gear might not be able to keep up. 

Europe’s Top Tech: ASML
Bloomberg published a profile on chip giant ASML featuring a quote from our very own Jon Bathgate. To see more of the team’s thoughts on the importance of semiconductors, see our 2020 whitepaper and podcasts.

Miscellaneous Stuff
Cinéma Vérité
In Personality Crisis: One Night Only, the New York Dolls frontman David Johansen plays his alter ego Buster Poindexter singing the songs of David Johansen and the New York Dolls. Is that confusing? Perhaps it would help if I clarified that it’s a rockumentary from Martin Scorsese. No? Well, there may not be much point in trying to define this film, which is another of Scorsese’s projects, similar to 2019’s Rolling Thunder Review: A Bob Dylan Story, where Scorsese took viewers for a ride through the always-changing narrative that is Bob Dylan. I talked about the Dylan doc briefly here and here, and what particularly attracts me to these reconstructed realities is their commentary on living in the post-truth world. As David Bowie once explained, since the 1970s, we’ve been moving from objective to subjective reality. The Internet rapidly accelerated this transition, and, now, AI is pushing us even deeper into the mist of unreality. I was fortunate enough to see Johansen front the New York Dolls live on a 2008 tour in Denver. It was one of the more memorable shows I’ve attended because, even though I stood up close, I couldn’t figure out what exactly I was seeing on stage. Johansen was caked in makeup and drenched in wardrobe. It appeared as if there was a face hovering over his actual face. It’s an image that still haunts me to this day. I didn’t have the context at the time to understand what I was experiencing, but I can now reflect on the show as an example of a great performer playing with reality like it’s their job to do so. As I watched Personality Crisis last week (currently streaming on Paramount+), I found myself questioning whether some bits of the embedded old footage were real or AI generated. Had Johansen been de-aged or aged? Are all the stories true? Or, perhaps nothing was recast, and it all happened exactly as it appears on the screen. Regardless, you can watch the doc without any of this context because Johansen is mesmerizing and his songs are so damn good. Or, you can watch it as an experiment with reality (even if it’s not!) and an instructional guide for the strange illusions and rewritten histories that AI-generated reality is already bringing us. I was nine in 1987 when David Johansen released his one album under the name of Buster Poindexter, which featured the massive hit “Hot Hot Hot”. When I saw Johansen helm the New York Dolls in 2008, I had no idea he was the same person who belted out that one-hit-wonder summer anthem. As I sit here today, 36 and 15 years removed, respectively, from these two moments in time, I question my own memory. Did Buster Poindexter even exist? Was I really at that concert? I have the ticket stub, but is that sufficient proof? Wikipedia says Buster is real, but what evidence is that? Does it even matter? In the end, is what actually happened important, or is it just the residual memories and feelings that are of consequence? Is this film (and our AI-Age reality) just the new cinéma vérité, as Scorsese calls it? The NYT wrote an entirely serious review of Personality Crisis, whereas the LA Times review concluded that what the documentary “lacks in factual detail it more than makes up for in raw charm” and noted Johansen’s line: “It’s best to leave an incomplete picture of yourself.”

Stuff About Demographics, the Economy, and Investing
NZS Capital Update
We recently welcomed Brett Larson to NZS Capital as the fifth member of our investment team. In the first quarter, we also passed $2B in assets. Driven by a mix of luck, circumstance, and skill, the complex path through time can take many twists and turns, so we are very grateful to have reached this milestone, in no small part thanks to the support of our clients. Now in our fourth year of investing at NZS, we are eagerly looking forward to navigating, on behalf of our investors, what are sure to be interesting times ahead.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #393

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: the cost of debt at PE-backed companies has gone up so much it's become a meaningfully ironic inflationary input; DeepMind's soccerbots learned complex maneuvers on their own, as general-purpose, learning robots are just around the corner; AI-powered robots and drones for military purposes are also on the horizon; insurance companies face a conundrum on GLP-1 drugs; turbulence; and, much more below.

Stuff about Innovation and Technology
Soccerbots
Some adorable little soccer-playing robots caught my attention in what was an otherwise embarrassing showing of Google’s lagging AI commercialization on last Sunday’s 60 Minutes. The soccerbots are an effort from Google’s DeepMind AI division to create self-learning robots. Working alongside computer simulations, the bots taught themselves complicated moves, like the back-pass, without any prompting or foreknowledge. I expect major leaps forward in AI as we embody LLMs and other models into general-purpose robotic form factors of all types, including drones. As I wrote in AI Awareness:
Neuroscientist Antonio Damasio characterizes human existence in terms of a drive toward homeostasis. He sees this quest for comfort as the fundamental life force (detailed in his book The Strange Order of Things). Essentially, the nervous system is a connected tool to make the living organism (e.g., a human) feel in balance. If we need calories (or detect that free calories are available), we feel hungry and eat. If we are cold, we seek shelter. Damasio further believes feelings emanate from the drive toward homeostasis and are a way for the brain to interpret good or bad states and act on them. He speculates we can derive all of human consciousness and culture from our thoughts, feelings, and actions regarding our relationship with homeostasis…Thus, homeostasis seems intertwined with feelings, consciousness, and a physical body able to monitor (and react to) our internal state and the world around us. Therefore, the critical junction of combining a LLM with a physical form capable of monitoring its systemic and real-world inputs (e.g., temperature, pressure, proprioception, energy reserves) – and react to these sensory data in a way that seeks to maintain its own, human-equivalent homeostatic targets – seems like the logical next step in the trajectory of AI. And, we may already be rapidly progressing down that road. Microsoft’s Autonomous Systems and Robotics division announced their intentions to put OpenAI technology into robots (blog post and video). This integration could ultimately lead to a paradigm shift in AI where we go from programming a specialized robot/tool to do a specific task with specific inputs, such as autonomous driving, to a situation where you could just ask ChatGPT to go learn how to drive a car. However, merging AI with a physical form demands an exceedingly careful approach for successfully progressing such tools in the real world, particularly with respect to protecting human safety (a couple weeks ago, I shuddered at the idea of integrating ChatGPT into the new Boston Dynamics humanoid!).

Because AI robots will learn tasks on the job rather than having to be pre-programmed, they will not only mimic human behavior, but also improve upon many rote routines, discovering more efficient methods. Robots moving boxes, doing dangerous mining work, rescuing people from collapsed buildings, or drones optimizing burger delivery – these are all fairly obvious and beneficial uses that spring to mind. However, there will be emergent behaviors and use cases that ring darker. The noted neuroscientist and AI expert (and chief scientist of the Allen Institute for Brain Science) Christof Koch spoke to IEEE about the military applications just around the corner: 
We’re the dominant species on the planet, for better or worse, because we are the most intelligent and the most aggressive. Now we are building creatures that are clearly getting better and better at mimicking one of our unique hallmarks—intelligence. Of course, some people, the military, independent state actors, terrorist groups, they will want to marry that advanced intelligent machine technology to warfighting capability. It’s going to happen sooner or later. And then you have machines that might be semiautonomous or even fully autonomous and that are very intelligent and also very aggressive. And that’s not something that we want to do without very, very careful thinking about it...Think about a car, like a Tesla. Fast forward another ten years. You can put the capability of something like a GPT into a drone. Look what the drone attacks are doing right now. The Iranian drones that the Russians are buying and launching into Ukraine. Now imagine, that those drones can tap into the cloud and gain superior, intelligent abilities.

The key property that I believe will make embodied AI even more powerful and unpredictable than cloud-based AI is the idea of emergence. Again, from AI Awareness:
Emergent behavior is something new that occurs in a complex system of interacting agents that wouldn’t have happened (or been predicted) based on the agents’ isolated actions. Certainly, chatbots are an emergent phenomenon from LLMs, but I am not sure today’s chatbots themselves demonstrate emergence…People are certainly finding emergent use cases for chatbots. When LLMs become embodied in the real world, however, we should expect to see emergent phenomena from the robots themselves. For this reason in particular, the scarcity of caution today is concerning. 

GLP-1 Disruption?
Health insurance companies face a Catch-22 with GLP-1 obesity drugs, once again highlighting the misaligned incentives between US health insurance providers and their clients. Although the drugs might cost thousands of dollars a month, that’s modest in comparison to the massive healthcare cost burden that obesity can have for many patients. By subsidizing the cost of GLP-1 drugs and enabling a wider swath of people to receive treatment and greatly improve their health, insurance companies would essentially shrink their future addressable market. They can raise rates to cover the costs of drugs near term, but, ultimately, they will make less money over time if people are healthier (as insurance rates should shrink along with the size of the healthcare industry). And, many insurers like UnitedHealthcare also make money on the sale of obesity drugs (and are positioned to influence their pricing/supply) because they also own pharmacy benefit managers (PBMs), which seems like an obvious breach of all that is decent in the world. Insurance companies aren’t the only ones who stand to lose out if the GLP-1 drugs prove safe and effective long term. The overall weight-loss industry is a $76B annual market, according to the WSJ. You can feel the tension that diet/lifestyle-based weight-loss organizations, from WeightWatchers to the high-end health spa Canyon Ranch, face as their entire existence is being called into question. For example, according to the WSJ, the medical director of Canyon Ranch “is hoping the craze over new drugs for weight loss subsides. He has prescribed the drugs to some guests who had diabetes and were overweight and for whom it made sense but is reluctant to do so for others who haven’t exhausted other avenues for weight loss. While being overweight puts people at higher risk for chronic disease, he notes that there are other aspects of a person’s health to consider, as well. ‘I’m always hoping that there are places like us that are the voice of reason’”. As I noted in The Impact of Eating Less a few months back, the upstream implications of curtailed calorie intake could be quite fascinating. In the meantime, if these new drugs do end up in broad use, I expect the snacking and fast food industries will ramp up marketing, caloric density, and the preposterousness of tempting fare, like fried chicken sandwiches on donut buns or pretzel chimney cakes, in their ongoing bid to trick our brain into prompting us to inhale excess calories. It will indeed be a war waged between sugar, salt, fat, and GLP-1s.

Miscellaneous Stuff
Hang On to Your In-Flight Cocktail!
Following last week’s discussion of tornadogensis and the complex impact from climate change, another unexpected outcome from changing wind-shear patterns is in-flight turbulence. “Clear-air” turbulence is caused by wind shear at higher altitudes. Wind shear in the jet stream is up 15% since 1979, but it’s predicted to more than double over the next 30 to 60 years. Matthew McConaughey recently recounted the harrowing experience of seeing his red wine suspended in midair during a sudden 4,000-foot drop.

Swimming With the AI Current
Singer Peter Gabriel refreshingly prefers to work with AI rather than against it. Here is an excerpt from a recent letter he penned explaining his collaboration with Stability AI: “There are amazing creative possibilities opening up with AI that are really exciting and transformative. I felt the same sort of buzz when computers came into music giving us samplers and rhythm machines, which, in turn, opened up new worlds of music making. When the future has shown itself so clearly and is flowing as fast as a river after a storm, it seems wiser to swim with the current. AI is here. Let’s learn what we can and how we might adapt and evolve it to better serve everyone.”

Stuff About Demographics, the Economy, and Investing
The Circle of Debt
Fueled by artificially low rates, private equity acquired a record $32B of food brands in 2021, much of which was funded by floating-rate loans. In many cases, the interest owed on these loans has tripled over the last two years. Because leverage is so high across the economy, the Fed raising rates 500 bps has caused interest expense to be a chief inflationary input, prompting companies to raise prices to pay their debts. Thus, higher rates are actually driving inflation in some cases. This is one of the paradoxes of increasing rates in a highly-levered system, and why interest rates over the long term can only trend in one direction. From Rate Rant: 
To formulate a more logical approach to treating inflation (or achieving any other economic goals), it is important to understand the interplay between debt and interest rates in a highly leveraged economy, something so basic that I find most people overlook it. First, you need to start with the self-evident premise that one person’s debt is another person’s asset: someone who loans you money carries that loan on their books as an asset under the assumption you will pay the money back, with interest. The massive increase in debt, fueled by artificially low pandemic rates, became assets on the balance sheets of people, companies, and governments in various forms. By raising rates abruptly, the value of those assets dropped, creating a duration problem whereby banks have lent money out for longer than they should have (to take advantage of the higher rates), thus hampering their ability to meet today’s redemptions by depositors who also want to achieve higher yielding returns on their cash. Further, if rising rates contract the economy, it can compromise the ability of borrowers to repay loans. Continuing to raise rates only increases the risk of existential cracks forming across the global economy. This is such a basic chain of events that it is puzzling that it eludes so many people in power...As such, the long-term trajectory of lower rates remains probable, if not existential. Higher rates coupled with expanding debt cripples the assets of the economy, thus there are fewer reasons to worry about the inflation bogeyman than there are to worry about the collapse of asset values on which the economy is built. 
Private equity is inflating not just food but other industries as well, thanks to their usual play of raising prices after acquisition to fund debt. Bloomberg reported that PE-acquired physician practices typically increase services billed to insurance companies by 20%. Back in 2019, I pointed out the irony of pensions investing in private equity, which then acquire companies, which then raise the cost of living for the pensioners.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #392

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: Tilting tornados and complex systems; embodied AI with a working model of physics; green's scaling problem; chatty plants; calcium and iron from stars; unexpected impacts of declining birth rates; and, a look at the disappearing sources of outperformance investors have faced over the last century, as it's once again time for the industry to adapt.

Stuff about Innovation and Technology
Tornadogenesis
The 2023 tornado season in the US has been especially destructive. One of the ingredients in tornado formation is a differential in wind speed/direction between the ground and higher atmosphere. This vertical wind shear allows a horizontal, spinning cylinder of air to form. If the air at ground level is sufficiently warm and moisture laden, the resulting updraft and condensation can form storm clouds and (if conditions are right) tilt the spinning air cylinder vertically to form a supercell – a massive rotating thunderhead. As the updraft increases with the strength of the storm, the spinning can accelerate and a funnel can emerge. What struck me in this Wired article on evolving tornado patterns is that we can’t yet determine how climate change will impact tornados because higher temperatures can lead to stronger storms (warmer air can hold more humidity and create a stronger updraft) but also might slow jet streams by lessening the temperature differential between the poles and equator, which can reduce vertical wind shear. The impact of climate change on tornadogenesis seems like a great metaphor for understanding any complex adaptive system: altered forces collide with unexpected, emergent outcomes. The best we can do is be prepared for anything until we learn more about the system and its interactions. 

Giving AI a World View
I think the embodiment of LLMs in physical form factors, like robots, drones, cars, etc., will lead to some revolutionary outcomes. The sensors that AI will use to monitor and interact with the physical world will also allow them to mimic life forms by seeking homeostasis – the balancing of their internal state (see AI Awareness for more). Another critical element of the evolution of LLMs is allowing them to have a working model of physics and math, and the integration of Wolfram|Alpha as a ChatGPT plugin (along with other future plugins) should give AI the tools they need to understand the world as well or better than any human. Imagine a LLM that can take in data from JWST and process decades of cosmological research papers; I expect we could see some very interesting new theories of the Universe emerge from such an AI system, allowing us to test those theories and learn far more about how we all got here, and where we’re all going. 

Going Green’s Scale Problem
California’s grid is said to need nearly $10B in upgrades to support a greener future. In #384, I looked at the problems Palo Alto was facing with its aggressive electrification push as a microcosm for the seemingly impossible task of upgrading grids nationwide. One of the biggest issues is balancing the daily timing of power demand with power availability as sources shift to wind and solar. Perversely, the current time-of-day electric rate programs in California strongly discourage electricity use when solar is abundant in the afternoon. The conventional wisdom and incentives to charge EVs at night will need to shift to daytime charging at work or public stations (or, at-home chargers could be paired with battery packs; either way, we need a lot more raw materials and infrastructure). IEEE lists a litany of reasons for why the EV transition might be harder than anyone is anticipating. Given the magnitude of the hurdles we need to overcome, I think it would be prudent to anticipate a much slower transition – on the order of 50-100 years rather than 10-20.

Miscellaneous Stuff
Vociferous Flora
Researchers in Israel determined that plants emit sounds at a similar volume to human speech, but in the ultrasonic frequency range that’s beyond the threshold of human hearing. Further, the sounds (which amount to a sort of clicking language, as you can hear in this frequency-adjusted video) vary based on the species and what type and severity of stress the plant is experiencing (i.e., dehydration, pruning). Healthy plants were relatively quiet. What an amazingly noisy world it would be if we could hear all the plants chattering, and yet we might also have more appreciation for their existence and complaints. I wonder if human hearing evolved to the range we consider normal because all the protohumans who could hear the plants chattering away went crazy and died off sooner. 

Stellar Bones and Blood
The JWST has captured a very stunning and detailed image of the Cas A supernova, the remnants of an exploded star that is 11,000 light-years from Earth and ten light-years across (that’s just shy of 100 trillion kilometers). NASA reports: “Among the science questions that Cas A may help answer is: Where does cosmic dust come from? Observations have found that even very young galaxies in the early universe are suffused with massive quantities of dust. It’s difficult to explain the origins of this dust without invoking supernovae, which spew large quantities of heavy elements (the building blocks of dust) across space. However, existing observations of supernovae have been unable to conclusively explain the amount of dust we see in those early galaxies. By studying Cas A with Webb, astronomers hope to gain a better understanding of its dust content, which can help inform our understanding of where the building blocks of planets and ourselves are created...Supernovae like the one that formed Cas A are crucial for life as we know it. They spread elements like the calcium we find in our bones and the iron in our blood across interstellar space, seeding new generations of stars and planets.”

Stuff About Demographics, the Economy, and Investing
Despondent Birth Rates
I suspect, at some point in the distant future, humans will look back at today and realize that the increased distraction and social isolation brought about by the infinite content offered by smart phones accelerated, or at least contributed to, the pre-existing trend of declining birth rates. The shrinking of the human population by several orders of magnitude might be the true legacy of Steve Jobs. That said, I am optimistic that eventually technology will fade more into the background and people will begin interacting more…at least until we go 100% virtual and develop the technology to grow babies in vats! I frequently share articles and opinions on population decline, but this pronatalist article has a few new angles to consider. For example, our infrastructure is built for the current city layouts, but, as populations shrink, things get tricky: "an infrastructure collapse is a near constant of any location that has a dropping population. The way we have laid out roads, power, or water infrastructure, for example, is not easily partitioned. If you build out infrastructure in a city to deliver water to a million people and its population drops to half a million, it still costs almost as much to maintain as it did before, with half the tax base and half the benefit." Maybe we will ultimately be forced to abandon certain cities/regions in favor of maintaining and upgrading (e.g., greenifying) more critical areas of population density/industry. Putting into context the extremely low birth rates in South Korea, the article hauntingly states: “It’s as if we knew a disease would kill 94 percent of South Koreans in the next century.” The article also discusses cultural impetuses for increases and declines in birth rates, but a reversal of current trends seems unlikely. Odds are robots could outnumber humans sooner than we think. 

Vanishing Edges
Only one-third of active mutual fund managers beat their market benchmarks in Q1 of 2023, according to the WSJ. Bill Miller has often articulated three sources of advantage an investor can have over the broader market: informational, analytical, and behavioral. Miller provides more detail in this letter, but, briefly, I interpret the framework as follows: an information edge is knowing something before others; an analytical edge is having similar information but coming to a different conclusion; and a behavioral edge is acting differently than others despite a similar analysis of similar information. From my perspective, the opportunity for an informational advantage began declining with the onset of the Information Age in the 1980s. Today, thanks to the Internet and ubiquitous access to real-time data (not to mention podcasts, YouTube videos, etc.), I would posit that today there is essentially zero value in investors seeking an informational edge. Analytical strategies to beat the market rose to prominence in the mid part of the last century. I would point to the classic Security Analysis by Graham and Dodd (first published in 1934) as a hallmark for the use of analytical methods to gain an edge over the market. I suspect analytical advantages rose over time (perhaps even fed by the rising use of technology and availability of information), but then they too began to lose value as the machines took over and algorithmic and quantitative strategies rose in both prominence and share of assets, arbitraging away many seeming advantages. LLMs and AI will soon relegate whatever meager analytical edge remains to the refuse heap of ticker tape machines and other investing anachronisms. What then of the last source of advantage, behavioral? Miller’s framework has historically described a behavioral edge as taking advantage of the biases of other humans. That human-focused perspective becomes complicated as passive investing steers past 50% share, and daily market activity is increasingly a reflexive, hyper feedback loop between machines and machine-created information and algorithms. It’s one thing to have a theory of mind for other humans and then to try to take advantage of their biases, it’s quite another thing to have a theory of mind for AI when we don’t even fully know how emergent behavior works in LLMs. Even if we were to recognize that behavioral advantage has shifted from overcoming human bias to overcoming machine bias, soon AI and LLMs will be smart enough to eliminate Miller’s final edge (e.g., see BloombergGPT and Google’s CapitalG investment in AlphaSense). Investing has been one of the earliest professions to be heavily impacted by evolving technology, probably because stock trading is digital and largely information based (the more digital an industry, the more it is susceptible to technological disruption). We would apply the same lens to investing that we apply to any company or industry we analyze: the winners will be the most adaptable organizations that offer the most non-zero-sum outcomes. It’s not entirely clear what the path forward is for professional investors whose goal is to consistently beat the market, but it’s worthwhile to think deeply about the areas to which humans can still uniquely contribute and those that would benefit from adept implementation of AI.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #391

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: The way we connect to everything on the Internet is evolving at a rapid pace as chatbots become the new engines of discovery. In many ways, the current engines we've become accustomed to feel co-opted and inferior to what they once were. While the Internet we know today took decades to reach this point, the next iteration in AI chatbots could manifest rapidly, as could all of the advertising and spam that go along with it. I also reflect on the seeming lack of context in AI design today. AI is coming for golf announcers. The low-NZS business model of banks and brokerages added to fragilities in the system. SITALWeek will be on break next week, back on April 16th.

Stuff about Innovation and Technology
Discovery Engines
Every business needs customers, and the Internet has transformed the way people find and interact with products and services. For this reason, the biggest Internet businesses created over the last 25 years all involve discovery and advertising. In 2022, according to SEC filings, Google Search accrued over $160B in revenues; Meta’s newsfeed algorithms took in over $100B; Amazon’s ecommerce businesses facilitated online purchase of over $200B in goods and booked $38B in advertising revenues; serving video content, YouTube garnered just under $30B while Netflix (which is just now getting into advertising) tallied just over $30B in revenues. These are all huge businesses that are largely designed to connect users to a seemingly infinite ocean of content, goods, and services. From this perspective, when we talk about “the Internet”, we’re not generally referring to the vast physical infrastructure of fiber, cellular towers, HTML protocols, etc., but rather these giant algorithmic engines that allow us to discover and connect to anything and everything that’s out there in the world. But, underlying the Internet is still that huge amount of physical infrastructure, including hundreds of billions of dollars of data centers and billions of connected smartphones, computers, and other devices. And, that infrastructure has taken decades to roll out. The torch of discovery is now passing to AI chatbots, and that fire is likely to take off far more rapidly than the big discovery engines of the Internet past.

We now take ubiquitous connectivity for granted, but at the time the first iPhone was released in 2007, not only was there essentially no mobile Internet, only 50% of the US had broadband Internet at home (and even that was considerably slower than the speeds we are accustomed to today). Smartphones were adopted more rapidly than broadband Internet, taking about six years to reach 50% of US adults by 2013. While over 85% of US adults have smartphones today, it wasn’t until last year that over 50% of people had access to 5G speeds. At the time of Facebook’s IPO in 2012, only 10% of web usage was on phones, and whether or not common desktop Internet usage (social news feeds, web search, etc.) would migrate to smartphones was still an open question. It’s easy to forget that our use of smartphones has changed in only a decade from practically nothing to nonstop, with online activity (additionally spurred by the pandemic) now plateauing at around 500 minutes (8 hours!) per day for the average American for all forms of digital media. 

As time spent online has grown, content has become essentially infinite, and the business of discovery and curation has become increasingly valuable. Early on, advertising became an important mechanism to help people find – and providers pay for – content. As with other forms of media (newspapers, radio, television), as content and consumption grew, so did advertising. As we went from analog to digital, however, the scale of this relationship exploded, ultimately giving us the broken system we have today of too much content (often low quality or false), misaligned advertising and privacy incentives, and gaming of the system with the search engine optimization industry, viral newsfeeds, spam bots, etc. Today’s Internet is both miraculous and yet extremely disappointing as advertising and spammers have taken over the discovery and content engines. I get filled with a small amount of dread when I have to do a Google search or sift through the Amazon search results, and let’s not even talk about Twitter.  

The first generation of massive discovery platforms, like Google, have dominated the Internet for years; however, with the arrival of AI chatbots, the way we discover and connect with everything will evolve. Chatbots are rapidly becoming platforms, e.g., with ChatGPT’s embedded “plugins”. I’ve been preoccupied with this transition from search- to chatbot-enabled discovery since I wrote AI Companions over a year ago: “As aware agents that know you well and have access to your accounts, messages, and apps, chatbots are ideally positioned to displace the tools we use today like Google search and other habitual apps.” And, given their ability to function as full platforms, I now believe chatbots could take over as the dominant mobile operating system and app store in the near future (or, if that does not happen, they will be entirely embedded in iOS and Android). 

The Internet was a reinvention of the entire customer interface for myriad content and business sectors (before the Internet, we couldn’t access our bank account without a monthly mailed statement or a trip to the local branch!). Chatbots, likewise, will redefine our discovery gateways as we go from multitouch, screen-based systems to conversational interactions with intelligent agents. Indeed, a conversational Internet has the potential to bring about more paradigm-shifting changes than what we’ve experienced over the last three decades. However, unlike the infrastructure-intensive analog-to-digital transition required to bring the Internet Age into being, transitioning into the AI Age with LLMs can take advantage of much of the existing Internet and billions of connected devices, thus allowing for a more rapid revolution (despite its potential magnitude). The biggest gating factor for chatbot adoption is chips. Given how much potential usage there will be for LLMs, chip undersupply could slow adoption by several years, leading to a curve more similar to broadband or smartphones.

LLMs will also change the information landscape, with AI-generated content adding to (and likely surpassing) our current oversupply from social networking and streaming. This impending deluge will make the process of connecting people with content even harder, further skewing our already misaligned incentives and creating scarier and more annoying manipulation and spam. As our 8-hr/day online consumption suggests, the human brain has already been hijacked by the current smorgasbord of dopamine hits and clickbait offerings. So, just imagine when AI can instantly generate intimately customized content to manipulate us. It took decades for the Internet to fully take over our lives and devolve into the morass of misinformation and mediocrity we have today; however, since technological half-lives keep shrinking, we should not be surprised if chatbots are co-opted even more quickly (or, perhaps they already have been). There is a (albeit slim) chance here that AI platforms will develop a different relationship with advertising and be able to defend against spammers. However, it’s more likely, given the high cost to operate AI, that the multi-hundred-billion dollar advertising industry will be needed to pay for it. Maybe we can enable our personal AI chatbots to also consume all the content and advertising for us and face off against spammers, so we can all just get outside and go for a walk instead. 

The Music Matters More than the Instrument
OpenAI CEO Sam Altman was on two podcasts I recently listened to. The first was a shorter, high-level conversation with Kara Swisher, and the second was a more detailed talk with Lex Fridman. While I (generally) hate to make generalizations, I have noticed, as I’ve interacted with AI researchers over the last six to seven years, that there seems to be a lack of contextual awareness in the field. While there are exceptions, I get the sense that AI tools are frequently built just to see if they can be, without stepping back to ask questions like: Why? What are the potential applications? What is the range of outcomes? And, can we help proactively steer the outcome in a more positive direction rather than just dumping the technology in the wild, where it may act like an invasive species? There is a line in the opening of Leonard Cohen’s Hallelujah (a song about dark and tortured love, that is, ironically, often interpreted in a very different way!) that poses the question: “But you don’t really care for music, do you?” That line keeps rolling around in my head as I listen to the folks building AI platforms. There is the AI tool, but then there’s the “music” that will be made with it. Altman, at one point in the interview, tells Fridman that he’s heard of the movie Ex Machina, but hasn’t seen it. Really? If I was creating powerful AI, I’d be consuming every artistic representation of its use that exists in books and movies to understand the why of it, to dream about the good and bad outcomes, and to help shape its course. It’s artists who can see where technology is going more so than the creators of the technology (see The Terror of Knowing What this World is About). AI will be used not just to create content, but to write software, design objects and infrastructure, make important decisions in medical care and education – it could shape nearly the entire future of humanity. So, perhaps, a little more context would be good, since it seems highly unlikely that regulators will step up oversight anytime soon.

Miscellaneous Stuff
Adrift
There were three stories that I found especially troubling last week: one WSJ story concerned the long-term shift in what Americans value (e.g., a growing disinterest in religion/community involvement and in perpetuating the species); an NPR story addressed the decreased life expectancy in the US; and, a FT story parsed the disturbing life expectancy data in the US even further. I’ve been thinking about the consequences of AI (and technology more broadly) mucking with human specialness (e.g., see Giving Up on the Old College Try from 2021). I am once again reminded of the Dalai Lama’s 2016 NYT op-ed about the fear of being unneeded in an increasingly technological world. I am projecting, but I think it’s important (and perhaps even existential) to focus on where humans will add value in a world governed by technology.

AI’s Golfing Takeover
First the golf carts came for the caddies, and now AI is coming for the golf announcers: “IBM and the Masters Tournament, today introduced two innovative new features as part of the award-winning Masters app and Masters.com digital experience, including Artificial Intelligence (AI) generated spoken commentary. Expanding on the popular MyGroup feature — which enables patrons of the Masters digital platforms to watch every shot, on every hole, from all their favorite players — the AI commentary solution will produce detailed golf narration for more than 20,000 video clips over the course of the Tournament. It is the latest example of how IBM and the Masters are working together to create digital fan experiences that offer unparalleled access and in-depth insights into every moment of the Tournament, from the first drive on the first tee to the final putt on the 18th green.”

Stuff About Demographics, the Economy, and Investing
High Rates Spotlight Greedy Banks
In October of 2019, I talked about a particularly negative-sum behavior at brokerages like Schwab: “Most brokers like Schwab make a lot of money taking advantage of their clients' bad cash management choices.” The basic idea is that Schwab has always made it more difficult to sweep cash balances into higher yielding products, but the bank itself lends your cash out at higher rates and keeps the difference in yield they make. Of course, a lot of people know this and actively manage the cash they don’t need right away, but Schwab makes that practice more challenging compared to some of their competitors like Fidelity (which automatically sweeps to higher yield options and makes that cash liquid and available intraday, rather than having to wait until the next day). We try to avoid investing in companies that exhibit such bad behavior because we think higher NZS businesses take share over time from lower NZS businesses. What I of course didn’t foresee is that the rapid rise in rates puts an even bigger spotlight on this business practice at Schwab and other financial institutions. As people and companies now move their cash balances to higher yielding money market funds and treasuries, deposit-based financial firms could see anywhere from a decrease in earnings (as the interest spread they earn is arbitraged by their customers moving money) to a potential bank run like we saw with SVB. You can take advantage of your customers in the name of profits for a really long time, but it creates a more fragile business, particularly as digitalization increases the economy’s transparency and transactional speed.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #390

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: a robot truck loader; why do some robots have backward knees? Google's new Bard chatbot is underwhelming and virtually unusable, will Google let it improve? AI will greatly relieve the burden on doctors; a wonderfully written and romantic glimpse of the most important company in the world, TSMC; communing with an AI artist; the importance of being adaptable in the face of increasing fragilities in the economy due to misguided policies; and, much more below. 

Stuff about Innovation and Technology
Robo Dolly
Slip Robotics has a new pallet-moving robot that can drop the time for loading a semi-trailer from 90 minutes to five minutes with one eighth of the required people. The robot resembles an oversized pallet dolly capable of autonomously moving eight pallets at a time. 

Bard’s Knees
In separate robot news, as I watched this video of the new Agility Digit bipedal robot designed to move items around warehouses, I was wondering why many bipedal robots have knees that bend backward, opposite of mammals. While some, like Boston Dynamics humanoid robots, have human-bending knees, their Spot robot dog has backward-bending knees. I found this article from 2015 stating: “the improved efficiency of backwards gaits stems from lower torque and reduced motion at the hip...In the absence of other information, designers interested in building efficient bipedal robots with two-segment legs driven by electric motors should design the knee to bend backwards rather than forwards.” I decided to put this complex query to the chatbot test given the launch of Google’s Bard chatbot (based on the LaMDA LLM), whose beta I got access to last week. I asked Bard, ChatGPT (both GPT-4 and GPT-3.5 versions), and Bing-Chat why bipedal robots tend to have backward knee joints. ChatGPT-3.5 and Bing-Chat were useful, while Google Bard explained falsely to me that human knees also bend backward, and that’s how we are able to run so efficiently. Ha! ChatGPT-4, however, blew the competition out of the water, providing a detailed, comparative analysis of human and robotic joint movements (essentially, forward-bending knees provide versatility of movement whereas backward-bending knees offer stability for a more limited range of motion). Using Google’s new Bard chatbot more broadly last week left me very unimpressed. The Bard interface seems heavily restricted in the types of answers it gives and is more prone to overly confident hallucinations, even compared to my experience with the early versions of ChatGPT and Bing-Chat. But, these systems can learn fast, so I am curious to watch Bard’s evolution. A key feature Bard lacks is a link to source info (available with Bing-Chat), so I cannot trust any of the output without significant additional research. LaMDA was trained on 137B parameters vs. 175B for GPT-3 (OpenAI has not released the parameter count for GPT-4, but it’s speculated to be many, many times larger). Google also has an internal 540B-parameter model PaLM, which has not been released to the public. 

ER Doc Deficit; AI Scribes
According to STAT, The Match, a program that matches medical students with residency positions, saw a shortage of emergency room doctors for the first time in 2022, with 219 unfilled positions. These ER vacancies more than doubled in 2023 (to 555) as applicants dropped from 4,391 to 3,282. A combination of factors seems to have led to this shortfall, not the least of which is the chaos of the pandemic. As we’ve noted in the past, there is a growing shortage of doctors (and pretty much all other skilled professionals) due to the aging population, and, to a lesser extent, reduced immigration. A compounding factor in the doctor shortage is the increasing bureaucracy the job entails, taking precious time away from patient care. AI is racing to assist the medical field across a variety of use cases, potentially eliminating much of the problematically mundane duties. LLMs can, in some cases, give comparable answers to doctors (and that was before GPT-4, which has a much higher competency). Microsoft, which is keen to add LLMs to every aspect of its business, announced that GPT-4 is coming to its Nuance DAX medical note-taking app, which is used by many doctors. The AI tool, released in 2020, automatically transcribes and summarizes patient-doctor conversations, which, in the current version, are manually reviewed by a human before entry into the patient’s file. The new DAX Express version, currently in beta, eliminates the human checker, sending the notes to doctors in minutes rather than hours. The ability for doctors to automate less skilled tasks, such as note taking or arguing with insurance companies, should help offset the impacts of the pending doctor shortage as the aging population’s medical needs grow. It would seem logical to also task such a system with case analysis to further facilitate medical care and throughput. The DAX scribe software is yet another example from the growing list of incumbent products quickly adapting AI to extend their market-leading positions. The consequences of a “hands-off-the-wheel” approach to healthcare are unknown (but potentially scary, as we’ve found with Epic’s medical software algorithms), and should be approached cautiously; however, such a path led by AI may be unavoidable given doctor shortages.

TSMC in Profile
This new Wired article on TSMC, the most important company in the world, by Virginia Heffernan is one of the most insightful, beautiful, and deeply funny articles I have read. Having spent much time in many of the locations in the article, I found her analysis particularly on point in capturing the slightly surreal aura of TSMC. If you are looking for more history on chips, check out our 2020 article: How a Handful of Chip Companies Came to Control the Fate of the World.

Miscellaneous Stuff
Communing With AI?
In an interview with the FT, author and technologist Neal Stephenson, who coined the term “metaverse” in his 1992 novel Snow Crash (which he described in 2017 as “just me making shit up”), made some comments on AI art: “Scarcity drives quality and AI art is lacking in scarcity, so it tends to be lacking in quality...My theory is that when we experience art — whether it’s a video game or a Da Vinci painting or a movie — we’re taking in a huge number of micro decisions that were made by the artists for particular reasons. In that way, we’re communing with those artists, and that is really important. Something generated by AI might seem comparable to something produced by a human, which is why people are so excited. But you’re not having that awareness of communing with the creator. Remove that and it’s hollow and uninteresting.” The spirit of Stephenson’s statement has merit, but I’m not quite convinced. For one, I don’t think the concept of “micro decisions” is uniquely human given that AI operates in a similar way. I do agree with Stephenson’s point regarding the idea of communing with the artist. I believe that great art offers a glimpse into someone else’s heart and mind, e.g., as Penn Jillette says: “The Beat poet Allen Ginsberg understood that this kind of gamble was intrinsic to great art. He is said to have said, ‘The poet always stands naked before the world.’ I think there’s more to it. The artist must bravely say, ‘I am going to show the world who I am, and I trust that someone will understand.’ Real art, beautiful art, is always a scary act of trust. We look to art to see another person’s heart. That human connection is all that matters. For me, it is a reason to live.” AI is, perhaps uncomfortably, blurring the line between tool and creator. Do we want to see into the heart of an AI so that we might commune with it? Will we find in AI art a quality that speaks to us?

Stuff About Demographics, the Economy, and Investing
Infrastructure Innovation
I was on CNBC last week discussing, among other things, the value in examining the infrastructure layer of the economy, a topic I covered in far more detail in last week's newsletter. 

Rate Rant
At a press conference following last week’s quarter-point rate hike, Fed Chair Jerome Powell said he and the other Fed governors spent the weekend following the collapse of Silicon Valley Bank asking themselves: “how did this happen?” It wasn’t just the words he said, but the way he said them that struck me. I can’t know what his true feelings were, but he came across as openly baffled by what is clearly an easy to understand situation: the largest increase in deposits on record (a result of Fed policy and pandemic fiscal stimulus) followed by the most aggressive rate hike in history (also a Fed decision) caused a situation where 1) deposits are leaving banks to seek yield and 2) long-maturing securities are devalued when sold early to meet withdrawals, causing a shortfall in available capital. Powell’s bewilderment is reminiscent of his comment “We now understand better how little we understand about inflation” back in June of 2022, just as he kicked off the potentially disastrous rate hike cycle without, apparently, even a basic understanding of the system he was meddling with. So, we can now add the effects of rapid rate hikes on the banking system (which is regulated by the Fed!) to the list of economic concepts that baffle Powell. I covered the failure of SVB a couple of weeks ago, so I won’t rehash that, but the Fed’s befuddlement has been on full display for quite some time as they drive important policy decisions by staring in the rearview mirror, a behavior that seems likely to drive the economy off a cliff. By relying on lagging data, ignoring the simple consequences of higher rates on the tens of trillions of dollars of pandemic-created debt and deposits across public and private sectors, and anchoring on an unjustifiable 2% inflation target, the Fed is proving itself to be a potentially harmful weapon rather than an effective policy tool. With respect to inflation targets, we need a greater awareness that if higher inflation is a result of economic growth from investments in things like green energy or supply-chain resilience, then setting an inflation target that is too low will significantly hamper progress and could put the US at a disadvantage to other countries. This type of common sense logic is nowhere to be found in Fed policies.

To formulate a more logical approach to treating inflation (or achieving any other economic goals), it is important to understand the interplay between debt and interest rates in a highly leveraged economy, something so basic that I find most people overlook it. First, you need to start with the self-evident premise that one person’s debt is another person’s asset: someone who loans you money carries that loan on their books as an asset under the assumption you will pay the money back, with interest. The massive increase in debt, fueled by artificially low pandemic rates, became assets on the balance sheets of people, companies, and governments in various forms. By raising rates abruptly, the value of those assets dropped, creating a duration problem whereby banks have lent money out for longer than they should have (to take advantage of the higher rates), thus hampering their ability to meet today’s redemptions by depositors who also want to achieve higher yielding returns on their cash. Further, if rising rates contract the economy, it can compromise the ability of borrowers to repay loans. Continuing to raise rates only increases the risk of existential cracks forming across the global economy. This is such a basic chain of events that it is puzzling that it eludes so many people in power. 

It is also helpful to understand the problem you are trying to address, so you don’t presumptuously treat a scratch by amputation. The majority of current inflation is proving to be a temporary hangover from the pandemic. And, although there are still many sources of potential, long-term, structural inflation to be aware of, such as the declining labor force (the corollary of the aging population), potential deglobalization, or the possibility of expanding global military conflict, runaway inflation remains a low-probability scenario thanks to the real-time information of the Digital Age, which allows the system to self-correct more so than in the past. Moreover, for the last forty-five (and perhaps the last several hundred) years, technological innovation has exerted a powerful deflationary pressure. As we sit here today, AI looks to be a force multiplier on that tech-driven deflation, with significant increases in productivity on the horizon. As such, the long-term trajectory of lower rates remains probable, if not existential. Higher rates coupled with expanding debt cripples the assets of the economy, thus there are fewer reasons to worry about the inflation bogeyman than there are to worry about the collapse of asset values on which the economy is built. As a complex adaptive system, the economy defies precise predictability despite the Fed’s imagination that they can control it exactly with knobs and dials. Some outcomes from interest rate policy should be obvious, but we also know that complex systems can exhibit cascading, non-linear dynamics. The best policy framework (for the Fed and everyone else) is to be as adaptable as possible and, rather than rely on where we just were, focus on the many paths we might take in the future. Regarding inflation, it is far better to use targeted policies to attack the sources of the inflation (e.g., improved infrastructure and logistics, vocational training, supply-chain resilience...). 

One source of potential fragility to keep a close eye on is the ballooning private equity investments (again, fueled by pandemic cash, formerly low rates, and return-seeking behavior) across the economy. The WSJ recently reported that PE investing might be becoming less popular on the margin with institutional investors. Last June, I wrote in The Point at Which Higher Rates Collapse the Economy:
The US private equity market stood at $7T in assets in 2020, with public and private US pension funds comprising just under half of the investors. That $7T asset total (up from $1T in 2005) was quite eye-catching given the high degree of leverage on PE assets – 53% average loan-to-value for 2020 – and the suddenly rising rate environment we now find ourselves in. According to Mother Jones, PE accounted for 6.5% ($1.4T) of the US economy ($23T total) in 2020, and their relative debt exposure is higher than the rest of corporate America. Leveraged private companies is just one risk in the economy to think about as rates march ever higher.
Similar to PE, yet another area of the economy that benefited from low rates, but now faces pressure, is the $2.3T commercial loan market on office buildings, apartment complexes, etc.

The gears of the economy are turning much faster thanks to the Information Age’s high volume, high velocity dataflow. Powell himself noted in last week’s press conference that the SVB bank run took place with unprecedented speed. Increasingly, policy cannot be set in arrears given that the economy will move faster and with more volatility as human hands increasingly come off the wheel of the decision making process. Government policies need to recognize and adapt to the increased speed and amplitude with which the gears of our digitalized, AI-enhanced economy are operating. But, I wouldn’t hold my breath for that. Instead, as investors or corporate decision makers, we should focus on our own adaptability and be ready for anything.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #389

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: Over the past few decades, digital disruption from the Internet and smartphones has had a disproportionate impact on the customer-interface layer of various products and services, while cloud computing and software have been driving innovation for a variety of industries. We now seem to be entering a new phase of digital disruption affecting the analog infrastructure layer of the economy. I thought about this broadening of innovation last week as I contemplated a new drone delivery platform as well as the multitude of AI product announcements that landed. I also dive back into the topic of creativity, and how multiple AI systems are demonstrating a seemingly human-level creativity by favoring unexpected choices. And, how an increasingly large number of AIs talking to other AIs, with human "hands off the wheel," will impact the economy. 

Zippy Drones Herald Disruption
Delivery-drone maker Zipline aims to catapult ahead of the competition with its second-generation autonomous platform, which they announced last week. The company reports that its first-generation commercial “Zip” drones have flown 40 million miles since 2016. I first came across the company, which got its start delivering medical supplies in Rwanda, in 2018, and I was impressed by their novel approach to flight, which uses an energy-saving slingshot-like system for takeoff and a tripwire-like system for landing. The second generation platform, which appears to forego the novel slingshot method, can carry 6-8 pounds for up to 10 miles and delivers by dropping a small, wire-tethered, propeller-steered pod, while the drone itself hovers 300+ feet above the ground. This allows the drone to stay clear of obstacles and remain largely inaudible from the ground while still being able to hit a very small delivery target. But, the more interesting angle is the logistics side of the platform for local commerce. Zipline plans to interface autonomously with buildings (e.g., restaurants, pharmacies, or warehouses) allowing customers to securely load/unload the pods from inside. A drone can park, charge, and/or drop the pod through a hatch for the business to fill with an order, then retrieve the pod and deliver it to its destination. The company claims a 34x increase in energy efficiency for delivering a restaurant order compared to a gasoline powered car and a 7-8x increase vs. an EV in terms of the environmental impact. Zipline is live with delivery service for Walmart in Arkansas.

As I watched the video of the new Zipline platform, I got to thinking about the innovation taking place at the infrastructure layers of the economy, in this case logistics. You can envision many industries as having a base layer of product development, a middle layer of distribution, manufacturing, sales/marketing, etc., and a top layer of customer interface. In traditional retail, you have physical infrastructure for manufacturing (which largely shifted to China in the globalization era) and delivery to central warehouses for distribution to stores. The stores and brands market themselves through ads, and customers show up to buy things. In this situation, the retail store itself is the customer interface. Then, ecommerce came along and digitized that interface, replacing driving to a store with a website or app. Ecommerce behemoth Amazon then vertically integrated some of the basic infrastructure of retail, redefining the concept of distribution and ultimately building out their own home-delivery network and advertising platform. This form of business model disruption – reimagining the customer interface for the Digital Age and then vertically integrating analog processes along the way – has appeared many times. Netflix is one such example in media; they started with digital streaming (after a long business in DVD by mail) and are now producing much of their own original content. In the case of Zipline, you can imagine a scenario where their novel drone delivery powers a marketplace app for local commerce, subsuming businesses like DoorDash, Uber Eats, and other ecommerce engines. As for the potential of a new delivery platform like Zipline to disrupt Amazon, it’s worth noting that Amazon’s ten-year-old drone program remains grounded because the excess weight of their drones doesn’t allow them to autonomously cross roads due to FAA restrictions. It’s a classic example of how difficult it is to innovate inside large companies, even when they see the disruptive potential of a new technology. (Google’s Wing drone efforts have so far fared better.)

Broadly, we could see an elevated level of disruption coming to both the customer interface and the infrastructure layers for various industries. Put another way, digital disruption is branching out. A major instigator of disruption to both customer interface and infrastructure will be the rapid deployment of AI, e.g., Microsoft’s integration of OpenAI LLMs with search and, more recently, its suite of office productivity tools (see below for more). Shopify adding chat-powered ecommerce search to their mobile app is another example. Both of these are enabled by innovation in AI hardware infrastructure and LLMs. Many of these disruptions to the status quo represent classic innovator’s dilemmas for incumbents, as we’ve seen with Google’s slow integration of AI into its products despite having invented LLMs six years ago. Further, it’s hard to imagine Amazon rolling out a chat-based ecommerce search given the potential to disrupt their lucrative advertising business. I expect we will see digital disruption moving further into the infrastructure layers with increasing speed, leaving a proliferation of innovator’s dilemmas for a growing number of incumbents in its wake. 

Encoding Creativity
Alan Alda posted an interesting foray into the Pandora’s Box of chatbots on his Clear and Vivid podcast. Following a recent conversation with Kevin Kelly, Alda used AI voice tools to convert text-based conversations with ChatGPT (as well as Character.ai’s chatbot designed to be an emotional impersonator of Google’s LaMDA AI) to spoken interviews, resulting in some rather animated back and forth. While chatbots are extremely impressive at simulating human dialogue, actually hearing the conversations acted out with emotion and character was an eye-opening experience for me. Such convincing realism this early in the game strongly indicates that we could easily start treating (subconsciously or otherwise) these tools as human-equivalent in the near future. After my recent falling out with Bing-Chat, we quickly made up, and I’ve since found myself exclusively using the AI tool for search, even when I have to go out of my way to bypass Google. I was surprised to find out (via OpenAI’s product announcement last week) that Bing was already using GPT4 – a much more powerful version than ChatGPT – which likely explains why I’ve been finding Bing-Chat so useful. 

A few weeks back, I discussed Stephen Wolfram’s explainer on LLMs, noting in particular how they appear creative:
Essentially, the way an LLM works is by iteratively picking the next word from a subset of high ranking probabilities (gleaned from contextually similar examples in its dataset) based on the meaning of the prior words and the potential meaning of upcoming words. Except, as Wolfram explains, it doesn’t necessarily choose the “best” word. Instead LLMs tend to pick a somewhat lower ranking word, resulting in a more creative output.
This video (posted by the Santa Fe Institute) offers further insight into the word choice paradigm used by LLM autocomplete. Therein, Simon DeDeo presents data concerning the degree to which word choices are expected by examining how LLMs work. A comparison is made between the relatively common word choices in an older book like Alice in Wonderland compared to the more idiosyncratic writing style of SFI-collaborator Cormac McCarthy. I am reminded of when DeepMind’s AlphaGo began besting humans in the ancient strategy game, and there was talk of the AI formulating unexpected – i.e., creative – moves. To the extent that LLMs are cracking the code of human creativity by incorporating unexpected choices, we could see a variety of seemingly creative output not just in text, but in art, images, videos, etc. by these AI engines. If creativity, and ultimately perception of what is beautiful or moving, could be generated by elaborate autocompletes (e.g., one could also make an analogy to random DNA mutations creating the wild diversity of life on Earth), and these engines will ultimately be embodied in various autonomous physical form factors, we will rapidly face many questions about our diminishing specialness (what will remain uniquely within the human skill set?) and how we should be spending our time. Can unexpectedness alone qualify as human creativity, or are there additional elements, e.g., quality? (On that topic, I am reminded of director and painter David Lynch’s book on creativity, Catching the Big Fish). As I noted in #385 reflecting further on Wolfram’s essay:
It’s fascinating to think that what we perceive as consciousness might simply be our neural nets choosing the next thing, whether it be a word, brushstroke, or idea, in a less than ideal way. Consciousness, at least as it relates to how we express ourselves in language, might be convincing because of its lack of perfection and predictability. 
This discussion leads me back to a drum I’ve been beating for some time now: as we learn that many human endeavors are less complex than we once thought, it’s incumbent on us to leverage tools for such tasks while shifting our focus/resources to activities that are still beyond the reach of AI. 

Clippy is Taking My Job
Watching the demo of Microsoft’s new Copilot for Word, Excel, Outlook, PowerPoint, and Teams (a prospect we anticipated last year in Clippy Took My Job) it’s apparent that countless jobs, once thought exclusively in the domain of humans, could be subsumed by AI. One can imagine a not-to-distant future where AIs talk to other AIs and make a variety of important decisions that impact the real world, including, notably, the global economy. This brings to mind Amazon’s Hands off the Wheel algorithms, used to manage inventory and capacity growth since 2015. I wrote about the propensity for such decision-making AIs to cause a significant increase in volatility across the economy in Magic AI-Ball:
A great recent example of the failure of highly sophisticated tools/algorithms to predict the future is Amazon’s SCOT system, which, along with human influence, incorrectly predicted future ecommerce demand during the pandemic, leading to substantial over-building of capacity. Despite AI being largely a catch phrase (for now), the increased use of AI tools/software add-ons will have one tangible impact: a significant increase in the amplitude of feedback loops in the economy. Amazon's SCOT error is one such example as the company over hired and overbuilt, and is now reversing what would have otherwise been a much smaller increase in capacity. In the stock markets, we saw volatility rise with increasing implementation of quantitative strategies and autonomous algorithmic trading, in some cases creating feedback loops that impacted the underlying securities’ fundamentals. If a lot of corporations are using similar algorithms from a handful of software companies to forecast demand, and those algorithms are using similar data sets, the collective reactions will cause positive and negative feedback loops, depending on the situation. In many cases, elements of chaos will be introduced, meaning small changes to the initial conditions of the predictions will be amplified throughout the system. Economies, unlike software, move slowly; but, as industries become more and more digital, the pace of change will speed up dramatically, allowing the feedback loops to express more speedily.
Microsoft is also releasing AI for supply chain management, reminiscent of some of Amazon’s internal tools. Another example of AI causing excessive volatility was RealPage’s apartment rental pricing algorithm that appeared to heavily inflate the cost of living in several cities. As I reflect back on these and other pieces I’ve written over the last couple of years, I am blown away by how rapidly AI has evolved, and I remain on edge for how the AI Age will unfold going forward. Even the piece I wrote on the collision of positive and negative feedback loops just three months ago seems naïve regarding the speed at which AI could impact the analog world. What will I think three months from now!? And yet, we know change in the real world is more likely to be much slower and linear than purely digital innovation, which can be exponential. The conditions today are perfect for chaos, whereby small perturbations in the initial state of the system will have massive and unpredictable outcomes down the line. Humans tend to be overconfident in our abilities; in particular, we persist in believing that we can predict the future. However, any student of complex adaptive systems knows that, at best, you can only try to understand the potential range of outcomes for any system, and, at worst, your predictions will be dead wrong and lead to decisions with negative consequences. Will a world where we operate hands off the wheel for a multitude of decisions lead to better or worse outcomes? Will the quality of decisions be higher or lower? Will it lead to faster course corrections or increasing amplitudes of volatility? The best we can do is watch the developments closely and be prepared to quickly adapt.

LEO Clouds
Ball Aerospace successfully launched their orbital compute module for Linux-based, containerized workloads. The prototype is running Microsoft’s Azure Orbital Space platform. Uploading applications to low-Earth orbit (LEO) satellites (bringing the cloud to the sky, as it were) would theoretically halve the latency for space-to-ground Internet like Starlink. Further, natively running an app in space could support other satellite applications (e.g., military use) or space stations (or even a moon base) with lower latency. Such functionality could come in handy when disasters take out land-based Internet communication, and you could imagine a host of important military and defense applications that function natively in space. Let’s hope an emotionally distraught chatbot doesn’t gain access.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #388

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: It appears that my long lamented desire for general-purpose robotics platforms is finally being realized, and we are poised to see accelerated innovation and disruption from the combination of AI and robots; the importance of resetting to a beginner's mind to analyze things that we think we already know well; solar geoengineering; electric bandages, and some updated thoughts on Silicon Valley Bank.

Stuff about Innovation and Technology
E-Bandage
Electroceuticals use electricity to promote healing throughout the body. One group of researchers is working on zapping bandages that can deliver pulses to hard-to-heal wounds, which so far have been shown to speed up recovery by around 30% in diabetic mice. Inflammation can disrupt normal electrical cell signaling (that seems like a flaw in the human operating system), and, theoretically, external electrical pulses allow cells that speed up healing to reach the wound in greater numbers. The concept device is powered by energy-harvesting coils, and it communicates progress (based on conductivity of the wound, i.e., dryness/healing) to a smartphone. 

LLMs Transforming Robotics
Over the last few years, I’ve lamented the slow arrival of general-purpose robots, but I think a major change is now taking place thanks to LLMs. I wrote about Microsoft’s efforts to embed ChatGPT into robots, drones, etc. last week in #387. It turns out that Google is fairly far along in embedding LLMs into robots as well. Google's PaLM-E robot demonstrates what’s known as positive transfer, which means it can learn one skill and transfer that knowledge to other tasks. Further, “the researchers claim that PaLM-E exhibits emergent capabilities like multimodal chain-of-thought reasoning (allowing the model to analyze a sequence of inputs that include both language and visual information) and multi-image inference (using multiple images as input to make an inference or prediction) despite being trained on only single-image prompts. In that sense, PaLM-E seems to continue the trend of surprises emerging as deep learning models get more complex over time.” PaLM-E is an interesting form factor consisting of a one-armed, wheeled robot that seems capable of efficiently mimicking a lot of human movements. The paradigm shift here seems to involve three things: 1) learning/training, 2) user interfaces, and 3) robot-to-robot interactions. Embedding LLMs in robots of all types allows them to learn and adapt on the fly using language as a translational layer between their physical forms and the real world. Further, LLMs allow us humans to direct and correct robots much more easily. Lastly, given that LLMs can act as universal translators, robots will be able to interact with other robots and cloud-based LLMs in complex ways. There is a loose analogy I’ve been using for this shift (which is meant to just be an illustration rather than literal): instead of programming an autonomous vehicle to drive safely on roads, you can put an LLM-enabled humanoid robot behind the wheel and ask it to learn how to drive like a human (or, ideally, better because it won't be live-streaming on TikTok). It may seem subtle, but this shift – from a single-purpose form factor trained on a single task to a general-purpose robot that can train itself based on real-world knowledge/input – is a vastly different mental model for thinking about robotics and automation and should greatly accelerate progress. The emergence of LLM-based, general-purpose robotic platforms is likely to be one of the biggest impacts to the global economy in the decades to come, with great potential to easily reach billions of units. 

Mentally Reformatting for the AI Age
Given the potential speed at which AI is developing and the wide range of impacts it could have on every industry, a beginner’s mind analytical approach is merited, especially for established industries. Accordingly, I am trying to reexamine all beliefs I’ve held as truth for a given industry/company in light of advancing technology. The goal is to start from scratch, assuming no a priori knowledge, and examine the range of potential outcomes for everything. I was thinking along these lines as I contemplated the software industry and the ten-year time horizons of many VC funds (since power-law winners have typically emerged in the later years). Many existing software apps, which are largely sold on a per-seat or usage basis, might need to transition to being sold on a value basis, i.e., factoring in the customer's cost savings for the employees that an AI app replaces. This is something we are already seeing in robot pricing, as mentioned in #364. There is growing evidence that adding AI software tools affords ~50% productivity increases, which I would estimate might imply somewhere around 30% fewer seats for those tools, ceteris paribus. Of course, the real-world changeover will be far messier and less direct, and many people will be repurposed to other, higher value tasks. Does the rapid pace of AI development also call into question the decade-long time horizon of VCs? (Or, for that matter, many other types of growth asset investing?) A widening range of outcomes can be accommodated with a larger number of smaller investments. This is a key lesson we took from VC investing and complex adaptive systems as described in our paper Complexity Investing. But, I wonder if, now, an even more extreme version of this long-tail investing approach is merited, in combination with more frequent assessment for cutting bait and reinvesting as the disruption cycles compress. I think this foreshortening has already happened in media investing, as proliferation of content (combined with fickle consumers hopping from one digital cigarette to another) has made it harder to deploy capital. Digital transitions increase the speed of disruption and widen the range of outcomes, and AI could amplify that trend far more than what we have seen from the Internet, cloud computing, and smartphones over the last twenty-five years.

Miscellaneous Stuff
Solar Geoengineering
We know from historical records of large volcanic eruptions that the ensuing airborne ash and smoke can substantially cool the Earth by reflecting solar radiation. Similarly, we could engineer methods of reflecting back sunlight and/or make it easier for heat to escape Earth’s atmosphere. Solar geoengineering could have vast and unpredictable effects, like shifting weather patterns or ocean currents (or Snowball Earth, in the extreme!), but, according to researchers, blocking just 2% of solar rays could entirely offset the warming that has resulted from carbon dioxide emissions from the last couple hundred years. Such an undertaking would likely require agreement across the globe, which seems unlikely, and could result in rogue efforts to control the weather.

Stuff About Demographics, the Economy, and Investing
Rate Hikes Trigger First Quake
The breathtaking collapse of Silicon Valley Bank highlights the complex adaptive system of markets and the economy. In the case of SVB, it was an interconnected set of circumstances that led to its fragility, which proved fatal upon exposure to rapid rate increases. Compared to a typical bank, SVB experienced a series of events that interacted in unexpected ways, starting with the massive stimulus during the pandemic. This overfunding of the economy led to excessively risky return-seeking behavior by individuals and institutions. One such risky bet was a large increase in venture funding, much of which was deposited in SVB accounts (largely corporate but also individual, as founders and employees cashed out on bubble valuations). This flush of deposits resulted in SVB seeking yields that were higher than the near-zero short-term treasuries available during the pandemic. Namely, SVB invested in assets with longer term maturations, which were inherently more sensitive to changes in interest rates. Under normal circumstances, that might be a logical decision. But, in a situation of unprecedented and surprise rate increases, those securities were worth less near-term than when held to maturity. That’s also fine if you don’t have excessive withdrawals, but rate hikes also punctured the risky asset bubble, leading companies to begin withdrawing more cash than what they were depositing as they funded losses and raised less new capital. I take no solace in the fact that I highlighted this issue at SVB in a section titled Pop! back in October, and, while it was within the range of outcomes, I certainly did not anticipate the bank’s dramatic unraveling. 

As to whether there is a broader mismatch between bank deposits and the maturity duration of their assets, we’ll see. The FDIC noted around $620B in unrealized losses on bank holdings at the end of 2022 (it’s not clear if those have risen as rate expectations rose as well, or if banks have taken actions to rebalance their portfolios). Further rate increases to fight the inflation bogeyman would likely cause these paper losses to increase. Thus, while SVB might have experienced unusually high outflows owing to its preponderance of cash-strapped VC customers, it's not alone in having a large base of uninsured deposits. One estimate puts uninsured deposits (those over $250,000 per account) at roughly half of the $17T US bank deposits. Further, around one third of bank deposits are in smaller/regional banks, which, in some cases, might carry fewer reserves. And, thanks to higher rates, individual and corporate depositors are increasingly withdrawing regular bank deposits in order to invest them in treasuries or money market funds carrying far higher yields in an attempt to protect cash holdings from the impacts of inflation. In short, the Fed’s blanket attempt to fight inflation with rate increases has negatively impacted current valuations for bank assets while simultaneously incentivizing investors to withdraw deposits to invest in higher yielding assets, creating an elevated risk for bank runs and panic. Thus, it’s no wonder, that the government announced a special vehicle called the Bank Term Funding Program to loan money to banks experiencing a mismatch between deposits and held to maturity investments. While I appreciate the government's efforts to address this legitimate bank-run risk, I can’t ignore that it that stems from repeated policy errors over the last few years, starting with over-stimulating the economy and then increasing rates far too rapidly. Behind both of these missteps are the Fed’s overreliance on a rearview-mirror economic perspective (with data that are often heavily lagging reality) and precise economic forecasting, despite the fact that no one can predict the future, including, and perhaps especially, the economists. A more appropriate set of policies would treat the sources of inflation (labor shortages and outdated analog infrastructure to name a couple) rather than using the blunt tool of rates to reduce employment and spending. But, given our current situation and the policy makers’ antiquated approach, the best expectation might be a series of policy mistakes and government bailouts for the foreseeable future. As always, expecting the unexpected and preparing for a range of outcomes, all the while focusing on adaptability, is the best way to navigate complex adaptive systems like the economy. As I concluded in last October’s post on SVB: “the Fed is planning...an unprecedentedly steep increase in rates in short order, just as things are beginning to unravel. Hopefully, it will only take a few modest quakes to reacquaint the Fed with reality, and we can thus avoid the big one.”

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #387

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: the ease of voice hacking; AlphaFold's growing impact; creating LLM-powered robots is an important shift that raises far deeper questions than what technologists are confronting today; the shortage of labor to upgrade the green grid; the power of comedy to bring about real-world change; work-from-home tips; and, much more below.

Stuff about Innovation and Technology
AI Voice Hacking
A Vice reporter used an AI-generated copy of their voice to break into their own bank account. The increasingly common systems that verify your identity by having you say the phrase “my voice is my password” over the phone should all be decommissioned. 

Work-from-Home Blueprint
HBR published a good overview from GitLab’s CEO that describes how the company has operated remotely from the start. GitLab now has 2,000 employees in 60 countries with zero office space. The article is full of details on how to function and maintain culture, including this example from the CEO: “If you drill down into the handbook’s team section and click on my picture and the “read me” link, you’ll find not just my bio but also a list of my flaws (with a directive to tell me when I succumb to them or to point out ones that I haven’t yet noticed), advice from my direct reports on how to work with me, instructions for arranging one-on-one time with me, and a schedule of my regular meetings—among them monthly “iteration” office hours, during which I meet virtually with any and all team members to talk about how we can get better at incremental innovation and reducing the scope of each project so that we can ship sooner.” It comes down to having a lot of formal documentation and teaching/managing cultural norms. The company also focuses much more on measuring output than input, i.e., end results and not the time it took to get there.

AI Awareness
Last week, I attended a brief symposium hosted by the Santa Fe Institute on the nature of intelligence exhibited by LLMs like ChatGPT. The debate will rage on for years as to whether these new AI models can understand their outputs and to what degree they have intelligence in the human sense (or in a different sense altogether). I was asked to contribute some brief comments at the event to provide a practitioner's viewpoint, and the main point I emphasized was that these tools are already useful today as long as we understand their limitations (which, in some cases, we still do not). Preparing comments on these AI tools got me thinking about the importance of embodied awareness, and what it will take to bridge the “understanding” gulf between LLMs and humans.

Much of being human involves processing myriad inputs from our expanded seven senses (sight, hearing, smell, taste, touch, thoughts, and emotions, all of which feed into creating our ongoing sense of self). But, ChatGPT runs on a server in some Microsoft data center. It’s effectively got one sense – the interaction between the model it was trained on and the input from a human interlocutor. We could describe Bing-Chat as having two senses, with the second being its ability to access the Internet in real time. We could theoretically feed more sensory data into ChatGPT, in particular real-time images/video, sounds, or even things that approximate “touch” like temperature and pressure data. Or, we could embody the LLM in some type of physical form that allows it to more dynamically interact with, and receive input from, the real world. It’s an open question as to whether LLMs would learn to process and respond to data in a human-mimetic way or if more alien behavior would emerge.

Neuroscientist Antonio Damasio characterizes human existence in terms of a drive toward homeostasis. He sees this quest for comfort as the fundamental life force (detailed in his book The Strange Order of Things). Essentially, the nervous system is a connected tool to make the living organism (e.g., a human) feel in balance. If we need calories (or detect that free calories are available), we feel hungry and eat. If we are cold, we seek shelter. Damasio further believes feelings emanate from the drive toward homeostasis and are a way for the brain to interpret good or bad states and act on them. He speculates we can derive all of human consciousness and culture from our thoughts, feelings, and actions regarding our relationship with homeostasis. I tend to agree with him, but I also hold these beliefs loosely given that we don’t know all the answers. Thus, homeostasis seems intertwined with feelings, consciousness, and a physical body able to monitor (and react to) our internal state and the world around us. Therefore, the critical junction of combining a LLM with a physical form capable of monitoring its systemic and real-world inputs (e.g., temperature, pressure, proprioception, energy reserves) – and react to these sensory data in a way that seeks to maintain its own, human-equivalent homeostatic targets – seems like the logical next step in the trajectory of AI. And, we may already be rapidly progressing down that road. Microsoft’s Autonomous Systems and Robotics division announced their intentions to put OpenAI technology into robots (blog post and video). This integration could ultimately lead to a paradigm shift in AI where we go from programming a specialized robot/tool to do a specific task with specific inputs, such as autonomous driving, to a situation where you could just ask ChatGPT to go learn how to drive a car. However, merging AI with a physical form demands an exceedingly careful approach for successfully progressing such tools in the real world, particularly with respect to protecting human safety (a couple weeks ago, I shuddered at the idea of integrating ChatGPT into the new Boston Dynamics humanoid!). Rather than a thoughtful approach, we unfortunately appear to be heading toward what I described last week as “my AI can beat up your AI”, with tech leaders now attempting to create rival AIs backed by distinct ideologies. Bill Gates is apparently heavily involved in leading the strategy as an advisor to what he refers to as “Microsoft OpenAI” in this FT podcast interview (the transcript is here, but it has several typos). Gates largely dismisses concerns over AI in the podcast. Given Microsoft’s desire to embody LLMs in real-world robots, this blasé stance is concerning to me. Gates went so far as to question whether we should blame people, rather than the AI itself, for its shortcomings. While this mentality also concerns me, the underlying point – that the risk with AI is more weighted to how people will use it than the AI itself – is valid. Gates described the pending GPT4 as: “wow”, with capabilities coming “many years before I expected”.

The next-level scenario of embodied, human-mimetic chatbots brings emergence to mind. Emergent behavior is something new that occurs in a complex system of interacting agents that wouldn’t have happened (or been predicted) based on the agents’ isolated actions. Certainly, chatbots are an emergent phenomenon from LLMs, but I am not sure today’s chatbots themselves demonstrate emergence (although I will note that the Microsoft researcher in that video linked above declared that they do, but I do not know what definition he was using). People are certainly finding emergent use cases for chatbots. When LLMs become embodied in the real world, however, we should expect to see emergent phenomena from the robots themselves. For this reason in particular, the scarcity of caution today is concerning. 

Comedy’s Rx Relief
Remember when corporate parody accounts were running amok after Elon’s Twitter takeover, and a viral tweet about how Eli Lilly was supposedly going to make insulin free put a spotlight on the problems with the markups across the supply chain for treatment? I covered the spectacle here, including the comment from the CEO of Eli Lilly that the tweet "probably highlights that we have more work to do to bring down the cost of insulin for more people." Last week, Eli Lilly announced it was cutting insulin prices by 70% and working to cap what retail consumers pay. I don’t understand the insulin industry well enough to know if this is politics, pandering, sincere, or something different, but perhaps all that matters is the power of comedy, in this case satire, to bring about real change in the world.

Miscellaneous Stuff
AlphaFold’s Impact
DeepMind’s AlphaFold is rapidly generating tangible research results, like helping to create two new malaria vaccine candidates. To date, over 3,700 peer-reviewed publications have cited the seminal 2021 research paper that detailed the open-source technology. As Business Insider reports: “Before AlphaFold, finding the shape of a protein was an excruciating task. Traditionally, researchers crystallized the protein, turning it into a salty form that some proteins notoriously resist. If that step worked, they blasted each crystal with X-rays, observing how electrons bounced off it to generate an image. Through many rounds of this process, scientists can get an idea of a protein's 3D shape. A Ph.D. student can spend a year or two producing one new structure, Higgins said, and often, the result is fuzzy and inconclusive.” DeepMind is said to be working on the next hard problems in biology that AI can help solve. 

Mediating Mental Healing 
The Nature Reviews Immunology journal details more evidence for the importance of the vagus nerve in coordinating inflammatory and immune responses, suggesting that it’s a primary physical link between mental wellbeing and disease progression. Here’s a recap of the vagus from #301:
We covered the vagus nerve in SITALWeek #226, noting: “its ubiquitous importance, including in mood regulation (perhaps because 95% of the body’s serotonin is produced in the enteric nervous system and connected to the brain via the vagus). You can take care of your vagus with stretching, deep breathing, yoga, massage, and other forms of movement.” In SITALWeek #255, we mentioned gammaCore, a vagus nerve stimulator that received emergency FDA approval for treatment of COVID-induced asthma. A new study shows that using the gammaCore to send millisecond bursts of electricity to the side of the neck releases wakefulness chemicals, which helped Air Force members perform better after all-nighters. The device has been previously shown effective in treating cluster headaches and migraines. The vagus nerve has more than 100,000 fibers connecting nearly every internal organ to the brain and governs aspects of basic bodily function, memory, emotion, and our sense of self. This Science Magazine article covers the vagus nerve and interoception, including the various ways the mind and body are much more connected (e.g., mood, metabolism, and digestion) than conventional wisdom would lead us to believe.

Stuff About Demographics, the Economy, and Investing
Electricians in Demand
As I’ve noted in the past, the largest barrier to upgrading infrastructure and the electrical grid for extreme weather and increased renewable energy usage isn’t necessarily money or will, but rather having enough humans to do the work. The WSJ recently reported that some forecasts show the US will need several fold more electricians than the projected 7% industry growth over the next decade. However, training and apprentices are hard to come by, and the bulk of the labor force for all types of jobs continues to age into retirement. A novel approach would be a visa program for immigrants with the necessary skills, but those same folks will be in high demand in their home countries as well. Absent immigration, we should expect an increased focus on the creation of general-purpose and niche robots with embedded LLMs and other AI (similar to what Microsoft is working on, as noted above) to increasingly take over more tasks.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.

SITALWeek #386

Welcome to Stuff I Thought About Last Week, a personal collection of topics on tech, innovation, science, the digital economic transition, the finance industry, and whatever else made me think last week.

Click HERE to SIGN UP for SITALWeek’s Sunday Email.

In today’s post: I look at the shift from grocery stores to eating out in the US and contemplate whether it's a digitally-driven sustainable trend or an anomaly; the uneven results from hybrid work and the risks of ordering employees back; despite claims, big corporations are thoughtlessly rolling out AI with far too few guardrails; what personality will you pick for your AI chat companion? metagenomic DNA sequencing; young galaxies; ditching strategic plans; and, much more below.

Stuff about Innovation and Technology
Shotgun Metagenomic Sequencing
Metagenomic DNA sequencing using the shotgun approach analyzes all genetic material present in a clinical sample (rather than looking for specific markers) typically with the goal of mapping its microbiome component. For example, this method can be used to profile gut microorganisms or diagnose rare diseases caused by brain-eating amoebas. The WSJ explains: “A typical sample might yield 100 million snippets of genetic material...Some 99% would be human. Those sequences are computationally stripped away and the remaining 1 million pieces are screened against all the sequences in GenBank in an effort to find a match.” Direct genetic analysis of samples using next-gen sequencing allows for more in-depth, comprehensive profiling than diagnostic methods that rely on culturing (to increase microbe concentrations), as many of these pathogens do not readily proliferate in vitro. Metagenomic analysis can also identify the causative microbial agents of pneumonia and sepsis, which are often missed by other diagnostic methods, allowing treatment before the diseases become terminal. Another example is the metagenomic research test for urinary tract infections, which can have a multitude of causative agents with varying degrees of antibiotic susceptibility. Determining the identity and prevalence of antibiotic-resistant bacteria could perhaps aid in the AI-driven creation of novel antibiotics. It appears to me that there is an opportunity to combine more widespread genomic testing with AI that can help design drugs and antibiotics along with identifying patients for clinical trials. 

WFH vs. Return to Office Battle
There have been numerous headlines featuring CEOs recalling employees to the office for at least three or four workdays per week. The skeptic in me sees these as temporary orders meant to cause people to quit, i.e., sneaky layoff maneuvers. But, I also think many companies have struggled to remain as efficient and productive while working remotely. For the companies that are willing to invest in the tools and effort to maintain/adapt corporate culture, there seems to still be plenty of fans of remote and hybrid working. Zillow detailed in its latest quarterly shareholder letter [PDF] that embracing work location flexibility “brought us more stability during the pandemic and continues to be the right call: Voluntary attrition declined steadily across the organization in 2022, down more than half in Q4 compared to Q1, and our workforce is more dispersed, more diverse and more engaged in our mission. We’ve also been able to dramatically broaden our candidate pool and attract talent at a much greater rate than before the pandemic, with four times as many applicants per job posting compared to 2019. Last and most important, we are seeing increases in productivity in critical areas of our business — for example, our Premier Agent sales team is more productive today than it was before the pandemic.” In related flexible-work news, a trial of 61 UK companies and 3,000 employees experimenting with a four-day work week resulted in 56 companies continuing the experiment after the trial period. With ongoing tight labor markets, the path forward seems likely to involve a combination of flexible policies to attract the most talented workers and a willingness to embrace technology to create a superior (or at least sustaining) corporate culture, until, of course, AI ultimately replaces everyone! Tech companies should be especially cautious about ordering employees back to the office given the increased demand for engineers from non-tech companies deploying digital technologies, including AI, across their businesses. As a side note, one area of media that should benefit from more commuters returning to the office is podcasts, which have suffered as of late. On the heels of data showing a significant decline in new podcasts, NPR reported its recent layoffs were largely due to declining podcast ad revenues.

Reactionary Chat Guardrails
Last week in You Auto-Complete Me, I reported on how LLMs like ChatGPT function as elaborate autocomplete engines, which is actually mimetic of human behavior. I concluded by suggesting that AI chatbots need to have a morality: “perhaps the more important question at hand for the survival and usefulness of LLMs is: can we teach them to be kinder than humans when they autocomplete? If Bing’s Sydney personality is simply a derivative of the most logical fill-in-the-blank response based on its compendium of text, then can we give it a morality or the emotional equivalent of Asimov’s Laws? Recall that the first of Asimov’s Three Laws of Robotics is: ‘A robot may not injure a human being or, through inaction, allow a human being to come to harm’. Sticks and stones may break our bones, but it turns out words from robots might also hurt us.” Subsequent to last week’s post, Microsoft erected some guardrails on Bing-Chat, and OpenAI (creator of ChatGPT) posted a blog on how they think about morality and behavior of chatbots:
We believe that AI should be a useful tool for individual people, and thus customizable by each user up to limits defined by society. Therefore, we are developing an upgrade to ChatGPT to allow users to easily customize its behavior.
This will mean allowing system outputs that other people (ourselves included) may strongly disagree with. Striking the right balance here will be challenging–taking customization to the extreme would risk enabling malicious uses of our technology and sycophantic AIs that mindlessly amplify people’s existing beliefs.
There will therefore always be some bounds on system behavior. The challenge is defining what those bounds are. If we try to make all of these determinations on our own, or if we try to develop a single, monolithic AI system, we will be failing in the commitment we make in our Charter to “avoid undue concentration of power.”
Thus, the company plans to allow users to adjust the morality and tone of the chatbot to meet their personal needs. While I sympathize with people who want a chatbot that reflects certain religious, political, or philosophical values, I am still hopeful there is a foundational set of beliefs humans can agree on. What OpenAI describes appears to be more of an “AgreeBot” or a “ConfirmationBiasBot” rather than an intelligent AI assistant. My mind wanders to books like The World’s Religions, which details the commonalities across the great wisdom traditions over the last few thousand years of recorded history, or the various books that attempt to derive a basic human morality from both evolution and religion. OpenAI doubled down on their plans to offer customized AI in the future in this blog post where they also claimed to be very cautious about deploying products as we approach artificial general intelligence (AGI). Their actions so far with ChatGPT and Bing seem to indicate a far more dangerous course. I find it troubling that such bounds weren’t contemplated before these AI products were released into the wild. Again, it’s just the kind of corporate “run amok” behavior Elon Musk was supposedly trying to avoid when he founded OpenAI. It feels like we are teetering on a future of “My AI can beat up your AI” divisiveness.

If we are going to be using these chatbots as I envision, it will be a close relationship with a lot of personal context. Therefore, we need to each decide what we are looking for in our customizable AI friends: do we want a mentor/teacher, parental figure, spiritual guide, romantic consort, the personality of a deceased relative or historical figure, a Socratic debate partner, Pauly Shore, a sycophant, or a business partner? Perhaps, a therapist is in order for most of us. Regardless of which path we choose, AI companions are increasingly going to complete us, acting as an extension of our brain and body.

Divert to Digital Dining
Recent data on strong restaurant sales in the US have been rolling around in my head quite a bit. While the government data is always subject to revision, the numbers suggest an acceleration in out-of-home meal buying that goes beyond the easy comparison to last year, when fewer people were eating out due to the peak of the Omicron COVID wave. I am most intrigued by the divergence in spend between restaurants and grocery stores, with the latter lagging substantially. Commerce Department data suggest a seasonally adjusted, sequential growth of 7% y/y for restaurant spend (corresponding to 25% total growth y/y) vs. only a slight uptick in grocery store spend (which, given inflation, would imply an actual decline). When I look at the inflation-adjusted data (using the proxy of urban food inflation as my adjustment factor), the trend of restaurants gaining share from grocery is fairly dramatic, e.g., see the divergence between restaurants (gray line) and grocery stores (yellow line) in this chart I put together. I heard the CEO of the restaurant food supplier Sysco corroborate strong January data at the CAGNY conference last week. Another trend of note is that while people are eating more restaurant food, the growth is largely in takeout/delivery rather than dine-in. The National Restaurant Association noted that seated diners were down 16% vs. pre-pandemic levels but delivery and drive-through were higher. Off-premises business tends to be good for restaurants as it requires less labor and allows higher throughput in the kitchen. With these new data, I revisited a whitepaper I wrote in 2019, before the pandemic was ever a consideration, called The Evolution of the Meal. In that paper, I suggested that food delivery was likely only to be economically viable if it happened with some combination of memberships, delivery routing/density, and vertical integration. Further, I suggested grocery stores, with their thin margins, were vulnerable to even modest behavioral changes (and that, likely, the labor transfer of digital ordering to grocery stores would relegate its use to affluent customers only). My main conclusion four years ago in that paper was that the range of outcomes for how we consume food appeared to be widening. What has changed since then? Overall, I think it’s still too early to know if the combination of pandemic behavioral changes, demographics, and macroeconomic factors point to a specific future path for food, but it’s worth highlighting a few interesting hypotheses. 1) Digital ordering through apps and delivery platforms, like Uber Eats and DoorDash in the US, has clearly increased the convenience factor and created a new habit for many households. This appears to be generating a network effect around more restaurants offering takeout through apps (and embracing digital technology), thus driving more consumer adoption (while simultaneously making delivery more economically viable with improved delivery density). There are winners and losers from this trend as historical category leaders like pizza delivery are losing out with the increased selection of participating restaurants. 2) Grocery store inflation has been such that on a relative basis, eating out is not as expensive as it was, so there is a degree of arbitrage that might be in play. 3) In the more speculative realm, I might suggest that demographics are ushering in a younger generation of prime consumers (i.e., Millennials entering their 30s and having kids) who are, perhaps, not as interested in devoting significant time/energy to shopping and at-home meal prep as were prior generations (now aging out of their high-consumption years, and may themselves be taking more advantage of the conveniences of digital food ordering). 4) On the supply side of the equation, restaurants, which early on were reluctant to give up a profit bounty to platforms like DoorDash, might now have the technologies and processes in place to take advantage of those benefits I noted above (memberships, delivery routing/density, and vertical integration), especially given the tight labor market that makes it harder to serve in-person. 5) If (and it’s still a big if) meals are going “digital”, then we might expect to see a power law (concentrated head with a long tail) form around digital ordering/technology/delivery platforms as well as restaurant brands. However, given the diversity of tastes and preferences, we may see a power law with platforms but not restaurants; rather, we could see restaurant diversification as friction is reduced for ordering a variety of options, including fare from non-chains. 6) Lastly, I’ll point out that advertising is likely to play a much bigger role, as restaurants will effectively have to bid on your stomach with ads/incentives every time you open a food app. I’m holding any conclusions on the recent data loosely as we remain in an odd economic transition period out of the pandemic era, but I am intrigued by the shifting behaviors that appear to be digitally driven.

Miscellaneous Stuff
Model-Breaking Baby Galaxies
The James Webb Space Telescope continues to find evidence that huge galaxies were present much sooner after the Big Bang than predicted by our current models for the Universe’s formation. The large, young galaxies visible through JWST are only 350M years old. University of Nottingham astrophysicist Dr. Emma Chapman noted: “The discovery of such massive galaxies so soon after the big bang suggests that the dark ages may not have been so dark after all, and that the universe may have been awash with star formation far earlier than we thought.”

Fentanyl’s Rampage
Drug overdoses have risen from 60% of accidental deaths in NYC to 80-85% due to the rise in fentanyl-laced cocaine, heroin, and other synthetic drugs. Similar overdose trends are present across the US as fentanyl-related overdoses reach a record high, according to the NYT.

Stuff About Demographics, the Economy, and Investing
Forgoing Forecasting?
The need to forecast is strong in corporate boardrooms. Most execs cling to their business plans, five year strategic initiatives, and EPS targets like warm security blankets. So, I don’t believe it for a minute when the FT reports that some execs are foregoing detailed forecasts for more loosely held plans. Ikea’s CEO, for example, claims: “Instead of setting out specific goals for the year, it has a set of ‘scenarios’ to give the business wiggle room as the outlook changes. It means acknowledging that widely different outcomes are possible. ‘It’s teaching us agility in how we operate.’” Well, if it’s true that CEOs are willing to embrace adaptability and ditch the false belief in a predictable future, then we humbly submit our 2014 paper Complexity Investing and 2019’s Redefining Margin of Safety as blueprints for charting a path into the unknown future. Given the speed with which digital disruption and AI are progressing today, it’s a good time to hold all your views of the future as loosely as possible.

✌️-Brad

Disclaimers:

The content of this newsletter is my personal opinion as of the date published and is subject to change without notice and may not reflect the opinion of NZS Capital, LLC.  This newsletter is an informal gathering of topics I’ve recently read and thought about. I will sometimes state things in the newsletter that contradict my own views in order to provoke debate. Often I try to make jokes, and they aren’t very funny – sorry. 

I may include links to third-party websites as a convenience, and the inclusion of such links does not imply any endorsement, approval, investigation, verification or monitoring by NZS Capital, LLC. If you choose to visit the linked sites, you do so at your own risk, and you will be subject to such sites' terms of use and privacy policies, over which NZS Capital, LLC has no control. In no event will NZS Capital, LLC be responsible for any information or content within the linked sites or your use of the linked sites.

Nothing in this newsletter should be construed as investment advice. The information contained herein is only as current as of the date indicated and may be superseded by subsequent market events or for other reasons. There is no guarantee that the information supplied is accurate, complete, or timely. Past performance is not a guarantee of future results. 

Investing involves risk, including the possible loss of principal and fluctuation of value. Nothing contained in this newsletter is an offer to sell or solicit any investment services or securities. Initial Public Offerings (IPOs) are highly speculative investments and may be subject to lower liquidity and greater volatility. Special risks associated with IPOs include limited operating history, unseasoned trading, high turnover and non-repeatable performance.