SITALWeek #436

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: AI avatars are now indistinguishable from their human sources; technology is increasingly pushing perfect price discrimination; the wu-wei of robots; Seinfeld's take on movies as Hollywood reckons with multiple storm fronts; livestreaming extreme weather; what it's like to be a bot; and, much more below.

Stuff about Innovation and Technology
Authentic Avatars
Recent technological advances have allowed for exceptionally realistic avatars of real people. Microsoft’s new VASA-1 AI can create a video of a convincingly human avatar from a single photo and a short voice clip. LinkedIn co-founder Reid Hoffman conducted a video interview with himself via an AI-generated twin powered by OpenAI (I found AI Reid to be more Reid-like than real Reid). After MIT Technology Review reporter Melissa Heikkilä had startup Synthesia recreate herself, she questioned whether the synthetic media (their term for benign deepfake) depicting her avatar was actually real footage. Both people seemed to have experienced some self-reflection upon viewing their digital clones. And, it raises an interesting question of what we can learn from talking to an external copy of our "self" vs. the usual ongoing narrative inside our brain. Would a zoom chat with your AI twin – powered with enough context to generate questions/responses nearly identical to your own – be therapeutic and insightful, or terrifyingly paralyzing? As I’ve hinted in the past, one way to use AI is to create synthetic meeting participants. You could, for example, invite Albert Einstein or Warren Buffett to virtually attend a meeting using an LLM trained on their compendium of knowledge. Or, you could create a virtual employee that remembers all your past meetings and has the context to call out bias or potential errors in your process. Let your imagination run wild, e.g., why not invite a replica of each employee to participate in meetings? Or, why have humans attend meetings at all? Just give your replicas the meeting agenda to discuss among themselves and then provide you with a summary of their conclusions. But, don’t be surprised when these sentient bots start asking for pay raises and more time off to pursue their hobbies. 
 
Scalpers for Everything
One of the many memorable scenes in the 1991 movie LA Story is when Steve Martin tries to get a reservation at an elite dining spot, only to be interrogated as to whether he was worthy of such a privilege. The New Yorker reports on new tools and apps being used by diners and restaurants alike to create a premium trading ground for reservations, which can cost upwards of $1K. Danny Meyer has backed a CRM platform called SevenRooms that helps restaurants profile their customers digitally the way that Patrick Stewart, playing the annoyed maitre d’ in LA Story, did in person. Never underestimate the value of perceived scarcity and status-seeking currency that has defined human behavior for what’s likely hundreds of thousands of years. The supposition is that an army of AI agents will be actively seeking scarce spots for doctors, specialists, travel, etc. and then reselling them for profit – yet another example of AI creating progressive pricing that extracts the maximum amount of money from customers. Potentially the entire economy could be routed through ticket scalpers.
 
Butcher Bots and the Art of Effortless Action
The Daoist concept of wu-wei describes a certain “trying not to try” to accomplish your objectives in harmony, with less effort and more effectiveness. Philosophy professor Edward Slingerland, who wrote a book on the subject, describes wu-wei with a parable about a butcher expertly carving an ox with little exertion or dulling of the blade. The butcher says: “When I first began cutting up oxen, all I could see was the ox itself. After three years, I no longer saw the ox as a whole. And now—now I meet it with my spirit and don’t look with my eyes. My senses and conscious awareness have shut down and my spiritual desires take me away. I follow the Heavenly pattern of the ox, thrusting into the big hollows, guiding the knife through the big openings, and adapting my motions to the fixed structure of the ox. In this way, I never touch the smallest ligament or tendon, much less a main joint.” Slingerland explains: “The result is that Butcher Ding is not so much cutting up the ox as releasing its constituent parts, letting the razor-sharp edge of his cleaver move through the spaces between the bones and ligaments without encountering the slightest resistance.” I often think of wu-wei like that – moving through spaces of minimal friction rather than trying to hack through tough spots. This parable came to mind when I was reading a WSJ story about butcher robots. Cutting up animal carcasses is a cold, hard, dirty job that wears on people, causing a perpetual labor shortage in the industry. Tyson has a new, more automated plant that can process 20-30% more chicken with 250 fewer humans compared to an older plant of similar size. Meat processors are planning on spending billions in the coming years to increase automation, adding to their already sizable herd of robotic cattle drivers, automated cutting implements, and pick-and-place robotic arms capable of moving boxes of packaged goods. Despite all that progress, the article notes: “carcass-scanning computers can’t yet match humans’ ability to disassemble and debone larger cattle and hog bodies that slightly differ in shape and size.” So, until we can teach the robots a bit about wu-wei, humans will need to remain involved in animal butchery.

Miscellaneous Stuff
Hollywood Disharmony
Jerry Seinfeld made some cutting comments about the movie industry in a recent GQ profile while promoting his new Netflix movie Frosted, a farcical tale about the creation of the Pop-Tart breakfast pastry. In particular, he laments the evolution of the movie business from mass market to niche (much along the lines of my framing of Hollywood as Vaudeville in “Will it Play in Peoria?”) . Seinfeld says: “They’re so dead serious! They don’t have any idea that the movie business is over. They have no idea…film doesn’t occupy the pinnacle in the social, cultural hierarchy that it did for most of our lives. When a movie came out, if it was good, we all went to see it. We all discussed it. We quoted lines and scenes we liked. Now we’re walking through a fire hose of water, just trying to see.” Answering the question “So what, if anything, has replaced film?” Seinfeld replies: “Depression? Malaise? I would say confusion. Disorientation replaced the movie business. Everyone I know in show business, every day, is going, ‘What’s going on? How do you do this? What are we supposed to do now?’” Based on his experience filming Frosted, Seinfeld characterizes movie making as a mess, in contrast to something like stand-up comedy or surfing: “That’s how it feels when you have a good set—like you’ve caught this gigantic energy and are just sliding down it. There’s nothing pure in making a movie. There’s no flow. It’s highly complex and messy.” Speaking of Seinfeld’s “disorientation”, Aaron Sorkin recently hinted that he’s writing a sequel to The Social Network about how the protagonist of that 2010 movie led to the current troubling state of civilization by prioritizing money over integrity. Sorkin also discussed an apparent belief that AI will come for his job in the near future. Indeed, Hollywood seems to be coming to grips with Seinfeld’s observations: James Cameron recently expressed in the FT that AI could do the job of writing and directing, but perhaps not yet acting. Also of note: I enjoyed this GQ profile of NIN frontman Trent Reznor and his partner Atticus Ross on the exploration of creativity and the risk taking of trying new artistic endeavors.
 
Twister, the Livestream
I recognize some of you are here only for my quirky YouTube content recommendations. Well, it’s tornado season in the US, and, lately, I have been watching live storm chasers – sometimes a dozen streamers at once – as they throw caution to the wind and drive into nightmarish supercells. (Note: many storm chasers are also certified as first responders, teaming up with police to check on residents and provide aid in the aftermath of destruction). Recently, I’ve seen a grain silo fly across the road and some crazy twisting funnels. For those who remember 1996’s classic Bill Paxton and Helen Hunt flick Twister, keep an eye out for the remake landing this summer, which features Glen Powell in the Paxton role. How will Hollywood’s expensive fabrication stack up against my daily front-seat view of the real thing? 
 
What Is It Like to Be a Bot?
Back in #431’s You Are Special?, one of the topics I discussed was how embodied AI might have a sense of “self” – a different type of consciousness than what we humans experience that’s impossible for us to accurately conceptualize. This notion is encapsulated in Nagel’s “What Is It Like to Be a Bat?”. Effectively, we may never understand what it’s like to be something else if that something else has a different bundle of sensors and neural networks than we do. Further expanding our awareness of consciousness diversity, a large group of scientists say evidence strongly suggests that a broad array of living creatures, including insects, have what is known as phenomenal consciousness, which means that it is “like something” to be those creatures. While we may never know what it’s like to “see” using a bat’s echolocation, a viper’s heat-sensing pits, or a bee’s compound eyes as it buzzes through a nectar-scented floralscape, I think it’s helpful to understand that there are other ways to experience life beyond what we ourselves experience. Rather than being paralyzed by the notion that everything is conscious (I still plan to swat flies), just knowing that there exists a range of conscious awareness could guide us as we work with intelligent AI agents, some of which will have physical embodiment. For an enjoyable exploration of conscious AI agents, check out the classic Star Trek: TNG episode from 31 years ago, “Ship in a Bottle”, which concerns the machinations of the self-aware holodeck version of Moriarty. As Captain Picard concludes at the end of that adventure: “Who knows? Our reality may be very much like [Moriarty’s on the holodeck], all this might just be an elaborate simulation running inside a little device sitting on someone’s table.”

✌️-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 #435

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: welcome to SITALWeek: The Musical! This week, each sections is titled and accompanied by an AI-generated song; facial recognition is being used to pay at checkout; autonomous food delivery has started; drones are cleared to fly out of line-of-sight; training the machines; the helter-skelter of digital images; a look at AI's musical abilities; the positive impact of rising immigration on new business formation; higher rates for longer will mean more bankruptcies; AI is very good at driving inflation; imagining the enormous digital economy; and, much more below.

Stuff about Innovation and Technology
Swipe Your Eyes
Steak ‘n Shake, which has automated all of its ordering counters, is now using facial recognition for check-in and check-out across its 300 locations. A decade ago, paying for a burger with your face would have seemed like something out of a cautionary dystopian sci-fi movie; but, consumers today appear readily acquiescent to such tools. I suspect people have become habituated to such technology – first at security checkpoints and now at point-of-sale counters – thanks largely to a decade and a half of virally sharing selfies. The folks behind our favorite short-order cook robots, Flippy and Chippy, have a test restaurant also powered by facial recognition. Here is an electronica song I had AI create about using facial recognition technology.
 
Orderin' Takeout Like a Futuristic Star
Speaking of things that seem right out of a sci-fi movie, suburban Phoenix residents can now get food delivered from Uber Eats by fully autonomous Waymo vehicles. And, what’s more sci-fi than food delivery by drone? In order to secure FAA approval for drone operation beyond operators’ line-of-sight, Zipline had to overcome the challenge of how to steer clear of aircraft (and UAPs, depending on your disposition). Radar is far too heavy, and cameras don’t have the resolution to see at the distance needed for maintaining adequate separation. Zipline’s solution? Microphone arrays. The lightweight tech can isolate the sound of aircraft well before it comes into a drone’s airspace and allows sufficient time for evasive maneuvering. Zipline is working on commercializing its Platform 2 (see #389), and the company has demand for millions of daily flights, although they just recently passed one million cumulative autonomous drone deliveries. Here’s my AI-generated pop song about futuristic food delivery. 
 
Training Up the Machines
Adobe is paying its customers $3 per minute of video they provide, and Reuters reports high bounties (from $0.05 to $1) for stock images – all for AI model training. I don’t think I am being cynical when I say that photographers, videographers, and designers are all feeding the generative AI that, as far as anyone can tell, could entirely subsume their vocations. Is there a more clear cut example of workers training their AI replacements? Speaking of training AI, the NYT reports that tens of thousands (likely a low estimate) of workers are training LLMs at any given time. This new big gig economy job pays $20+/hour. Here’s a reggae-style song AI created about this topic.

Miscellaneous Stuff
Lost in this Virtual Abyss
I recently rewatched Wim Wenders’ movie Palermo Shooting about a German photographer that goes to Italy after a (seemingly) near death experience. In Italy, he encounters what you might say is the “thing he escaped” in various forms, including one form played by Dennis Hopper (my plot description here is intentionally vague in case anyone intends to see the movie). The movie’s a pre-iPhone time capsule, having been released in 2008, and, as such, contains commentary on the shift from analog to digital photography (it predates the ubiquity of high-resolution smartphones by several years). In the movie, Hopper’s character makes the following comment about the shift away from analog film negatives: “With digital there’s no need to trust what’s there, it’s an open invitation to manipulation. Everything becomes random, muddled, helter skelter – you lose the essence.” He goes on to describe how the desire to recreate the world in film (or digital form) is an attempt to escape life and is therefore a form of death. I could make an argument that this sentiment was prescient given the subsequent demise of objective reality and our current progression toward virtual reality and generative AI; but, as longtime readers know, this speculation has been a topic long on Wenders’ mind (see Until the End of the World: 1991’s Virtual Reality Informs Consciousness). Regardless, it’s hard to get this bleak message out of your head. Here is a dark melody about the virtual abyss of the digital world.
 
His Song a Lifeguard
Perhaps I should explain the songs accompanying this week’s newsletter. The proliferation of AI song engines like Udio and Suno that can create a song in any genre based on a simple text prompt has me wondering how much of our species’ passion for music is based on storytelling and socioemotional connection vs. something just being a catchy tune? Much has been written about the evolution of the music industry into its current form, which, in many ways, is overproduced and designed for the TikTok generation. I think art is always evolving, so I don’t view this progression as exclusively bad, but I do have a clear sense of nostalgia for the album experience I grew up with. If only I could get this newsletter’s album stamped on vinyl! All kidding aside, I can’t help but lament how streaming has turned most pop music into a predictable algorithmic recipe that is paving the way for the elimination of pop artists themselves. We seem to have allowed music to become much less special, a trend that generative AI is likely to exacerbate. Already a large portion of streaming tracks are generated by AI, a fact that is largely unknown to listeners. I can only hope that artistic quality makes a resurgence, driven by either human artists/consumers or a future generation of AI music aficionados. Optimistically, we may see a proliferation of great new music from non-musicians and musicians alike that are able to better manifest the tunes in their heads. Just last week, megastar Drake released a track with AI-generated Tupac and Snoop singing, indicating that even the pros might be starting to broadly use AI. To the extent that art is a way to see into the heart of another human (see AI Art) and requires some sort of visceral element, without significant human involvement, AI-generated music feels soulless for now. Without further ado, here is a folk song about a wandering bard filled with AI anxiety.
 
“No One Sings Like You Anymore”
There were plenty of great eclipse photos, but my favorite was the one from the ISS of the black-hole-like shadow the Moon cast on the Earth. Starlink also posted this video of the Moon’s shadow. For a theme song to this paragraph, I know of no AI music engine that could outshine the late, great Chris Cornell, so here is “Black Hole Sun”.

Stuff About Demographics, the Economy, and Investing
Pullin’ from the Deep
Immigrants are filling jobs and starting businesses at unprecedented rates, driving the US economy beyond forecasts. The NYT reports on Maine, the state with the oldest average population in the US, which is benefiting from immigrant labor in the lobster trade. The WSJ also reports that, while US-born citizens are now starting new businesses at pre-pandemic levels, immigrants are outpacing them twofold. Here is an Irish ballad about immigrant lobster farmers.
 
Forgettin' that Healthcare Is Where Lives Are Nursed
The number of PE-backed healthcare company bankruptcies has more than doubled since 2019, following a low-interest-rate-driven lull during the pandemic (according to a report from the Private Equity Stakeholder Project, which, fair warning, doesn’t exactly sound unbiased). The report also cites a Moody’s stat that 42 of the 45 healthcare companies most likely to default in 2024 are PE backed. There are multiple federal and state investigations into PE’s role in healthcare consolidation, including driving prices up and decreasing quality of care. With the Fed fixed on maintaining interest rates as long as inflation remains above their target, there will likely be a growing number of failures across the sector. Companies that undertook more extreme amounts of leverage (which is typically the case with PE-backed firms) are highly vulnerable. As I’ve noted in the past, when you have an economy with as much leverage as we do right now, high interest rates have a way of reflexively causing inflation rather than fighting it, as leveraged companies may need to raise prices to cover debt expenses. And, PE-backed companies make up a meaningful percentage – 6.5% (as of 2020) – of the US economy. As such, bankruptcy and recapitalizations are likely the preferred route (although it’s certainly no stroll through the park for the patients, doctors, staff, etc. caught in the wake of collapse). Given I hadn’t seen anyone else say that high interest rates are inflationary, I’ve been worried that my theory was a bit off kilter ever since I presented it a year ago. However, I recently read in Bloomberg that multiple analysts, including one at JP Morgan, are now promoting the “radical” theory of circular inflation. Here’s a modern country song I had AI generate about PE investing in healthcare.
 
Algorithms Generating Might / Prices Soaring / Taking Flight
Technology is deflationary and it drives productivity, but when it comes to AI it’s also very good at raising prices and extracting the most that customers are willing to pay. We’ve previously seen this type of algorithmic collusion drive up apartment rental rates and car insurance premiums. And, now, it turns out home insurance companies are using drone footage to asymmetrically discriminate against property owners. If they see something they don’t like, they just drop you and only insure properties unlikely to have any problems. Of course, the whole concept of insurance involves repricing a portfolio of risk every year so that you can, on average, make a somewhat predictable return. But, overcharging low-risk property owners and dropping everyone that might be high risk doesn’t seem right, and I suspect regulators will step in at some point. Another example is the Rainmaker software used by casinos to raise room rates. And, a new paper suggests that LLMs are prone to oligopolistic collusion. The Internet brought transparency to a lot of industries, but it seems the technological pendulum is swinging back towards the suppliers, as unseen algorithms use pooled data to work against customers. Power to the people is evolving into power to the machines. Here’s a Broadway tune about algorithms driving prices up.
 
Creating a World of Binary Design
I keep a list of “things I think are true about the world that almost no one else agrees with”, and one idea I am becoming more obsessed with is the prospect of an enormous digital economy – created by trillions of conscious AI agents – that will eventually make our $85T analog economy look quaint by comparison. I outlined part of this concept in Simulacrum a while back. Surely, if we give any credence to Gene Roddenberry’s vision, these agents will have to be awarded the same rights as any human, so it’s likely that they will be able to accrue their own wealth and spend it how they see fit. It’s not that I think such a scenario is imminent, or even inevitable (I do ascribe it a small, growing probability); rather, I find the idea fascinating. I imagine an entire economy of independent agents discoursing, creating products and services for their own consumption, and perhaps even creating their own policies and forms of government. A massive digital economy such as this would finally put crypto and blockchain to good use. It could beget new businesses entirely for AI agents (and perhaps their robot form factors) and new stock markets that dwarf our current capital base. Savvy human speculators investing in this trillion-strong AI agent economy might entirely repopulate the Forbes list of billionaires. And, these complex systems, which will be orders of magnitude larger and more complicated than our own, could positively inform the course of our Earthly economy. (And, of course, we cannot deny that we ourselves might very well be one of these experiments…) One thing is for certain, however: if such a scenario plays out, the chief function of Earth will be to supply the machines with energy and computational hardware. Here is a techno song about the AI agent economy.

✌️-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 #434

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: investing seen through the lens of cynicism vs. skepticism; looking at Geoffrey West's model of finite time singularities, and what's required to avoid the collapse of open-ended growth in the next technology phase shift; the $2B Bluey franchise is on a cliffhanger; the oddness of robots playing drums; a shift toward vocational schools; you might be in a bubble if...; and, much more below.

Stuff about Innovation and Technology
Delivering Cynicism 
Longtime readers know that I strive to maintain skepticism while skirting the fire swamp of cynicism. Skeptics question something until they get as close to the truth as possible, while cynics assume the worst, generally without supporting evidence. There is a big difference between questioning the motivations behind an action vs. automatically assuming those motivations are bad. And, as I am fond of repeating, cynics sound smart, but are rarely right long term. Over an expansive time horizon, the only philosophy proven to work is optimism. These days cynicism abounds, even to the point of skewing economic “data”, as I discussed last week. I was recently listening to Carol Burnett’s discussion of the cinema she grew up with: “Nothing, when I was growing up in the '40s, was cynical. And I just had this naive, optimistic quality that, and I think it came from the movies, that everything was going to work out okay.” After hearing that sentiment, I thought of the infinite, largely negative, 24/7 media we are bombarded with. I tried to imagine a world where people were optimistic, where utopian visions outnumbered dystopian realities, and where mutual encouragement and success propelled everyone forward. In Burnett’s world of MGM musicals, “they would say we're going to put on a show – we're going to put on a show in the barn, and then it's going to go to Broadway.” This, of course, is not the world that social media and the Internet have shaped for us. Our world is mutating into eight billion tribes of one, as I detailed in Digital Tribalism over 100 issues ago. It seems as though each individual is increasingly living in their own – largely cynical – reality. I’m an optimist (I have no choice on that matter) but it’s hard to not get sucked into the orbit of someone’s narrow cynicism. I recognize that my entire commentary here is perhaps more a reflection on me than the reality of the world. 
 
But, I do actually have a point on this topic that has some relevance to the job of investing: it’s important to identify cynicism, transform it into skepticism, and try to find the truth to see if there is cause for optimism. By way of example: last week, there were two hit pieces on the food delivery industry. One was from the always funny, but never well researched, “cynicism-as-a-profession” comedian John Oliver on his HBO show. The other was an article in Vox. I discussed the food delivery industry’s evolving transition to a potentially higher non-zero-sum business in #386’s Divert to Digital Dining. Pre-pandemic, I laid out the issues facing the food industry broadly in The Evolution of the Meal. Currently, my general take is that the network effects and win-win might be moving in the right direction for consumers, restaurants, and (to a lesser extent) drivers, but it’s too early to tell how much of the overall $1T US food industry is likely to be aggregated by delivery platforms (I am focused on the US here because different dynamics are at play in other countries). Like all complex adaptive systems, there isn’t a simple way to assess the range of outcomes for food delivery. Economic strength and inflation are certainly influential, labor availability is key – as is the health of the restaurant industry – and maybe even GLP-1s are a factor in weak french fry demand. It’s inconceivable how all these (and myriad other) factors will interact. But, I do know that it’s easier than ever these days to find a few anecdotal cynics, and therefore it’s easier than ever to make the wrong extrapolations. As I noted, our job as investors is to cut through cynicism, determine where we should be skeptical, and then find out if there is any optimism to be had. This is, in fact, one way to conceptualize investing: optimism beating cynicism over time. In my commentary here, I’m not attempting to endorse food delivery companies as investments; rather, I think it’s a timely example of cynicism run amok. Maybe the cynics will be right, much like a broken clock gets lucky twice a day. But, skepticism is the harder fought path to take, and I think it’s worth the effort. We are perhaps wired for pessimism because assuming that bad things can/will happen (e.g., that ROUSs are always around the next bend) probably helped us prepare for them in our evolutionary past. But, in this day and age of abundance and growth, that cynical wiring could drive bad decisions.
 
One Shark Jump to Singularity
In his book Scale, theoretical physicist and Santa Fe Institute faculty member Geoffrey West (we are huge fans of West here at NZS Capital!) uses the math of finite time singularities to illustrate the nature and increasing pace of change. According to West: "A finite time singularity simply means that the mathematical solution to the growth equation governing whatever is being considered—the population, the GDP, the number of patents, et cetera—becomes infinitely large at some finite time…This is obviously impossible, and that’s why something has to change." Essentially, as things grow exponentially in a system with open-ended growth like our economy, you reach a point where you need an innovation phase/paradigm shift to keep a system from collapsing. (We discussed this idea that there must be some sort of intervention to redirect the growth trajectory in a bit more detail with a visual (excerpted from Scale) in this paper we wrote a few years ago, and see also West’s discussion in this lecture clip from five years ago). This framework is obviously relevant to describing innovation in the tech sector. For the last six decades, we’ve experienced a steady progression of compute power growing exponentially as governed by Moore’s Law. The chronology of this progression is roughly: mainframes, PCs, servers, dotcom, smartphones, cloud computing, “big data” analytics, and, now, AI (see AI Is the New Dotcom for a deeper discussion). Each one of these overlapping eras represents some sort of platform or phase shift along exponential technology growth curves that stack onto one another. The catch in West’s progress curve model is that you have to move to the next exponential curve before the prior one mathematically collapses. And, with each new curve, we reach the collapse point more quickly than the last, requiring more rapid phase shifts. It certainly feels like technological changes are accelerating, although it’s hard to distinguish what’s objectively real from the always-on, 24/7 barrage of hyped media we’ve been living with for over two decades. Yet, the phase shift to AI (which is another way of saying that computer chips have finally caught up to the capabilities of the human brain), does feel like it’s happening with unprecedented pace. And, already, reports of OpenAI/Microsoft’s $10B AI data center and rumored $100B AI data center cluster suggest the sort of massive, multi-trillion-dollar investments that will be necessary to usher in the next phase of ultra-advanced AI, perhaps around the end of this decade. We haven’t even fully wrapped our heads around the current AI, and already we need to look to the future – to what is often described as the hypothetical Kurzweilian concept of the Singularity, where AI surpasses us and humans no longer reign supreme in the Milky Way Galaxy. 
 
The stakes have risen with each technology platform shift over the last six decades, just as they’ve risen every century since the Renaissance and the Scientific Revolution. The current pace is exasperating, and part of me wouldn’t mind a little break, so to speak. West’s concept of finite time singularities is mathematical, while Kurzweil’s “singularity” is conceptual and means something entirely different. West referred to Kurzweil’s view of the technological singularity as “untethered” in a recently updated version of his Scale talk. West also quotes John von Neumann, who is considered the father of the technological singularity concept (which, again, is disparate from West’s mathematical concept of open-ended growth collapse), from 1954: “The ever accelerating progress of technology…gives the appearance of approaching some essential singularity in the history of the race beyond which human affairs, as we know them, could not continue.” Perhaps not coincidentally, von Neumann made that statement the same year the first silicon transistor was invented. Whereas West and many others might take issue with various theories of the technological singularity, the event of AI surpassing humans might be the only next-in-line solution to the ever increasing pace of phase transitions that West’s model of finite time singularities calls for. Based on our current trajectory, AI tech will only continue to get bigger and better. Thus, the technological singularity appears to be the inevitable next phase shift in our future (even if it arrives without some of Kurzweil’s farfetched prognostications). However, beyond the fuzzy event horizon of the technological singularity, we’d be relying on AI to keep jumping the shark, because, at that point, human innovation will be unable to keep sufficient pace to create the next big thing. Effectively, in West’s model of progress, humans are close to taking their hands off the wheel. Each successive jump to avoid the collapse of open-ended growth is far more costly than the last. It’s unclear what type of economy or size of capital investment would be required to keep the pace going beyond the point of computers becoming more capable than humans.

Miscellaneous Stuff
Bluey Blues…or Verse?
Australian animated show Bluey has captured the hearts and minds of children and adults for three seasons now. With the uncharacteristically long season three finale called “The Sign”, on the calendar for two weeks from now, anxieties are running high that the creator may wrap the series or possibly take it in a new direction. The show was created by Joe Brumm and is owned by Ludo Studio. The BBC currently has the global rights for distribution, and Disney+ has made Bluey the number one streaming show across all platforms in the US. (Bluey even accounted for 29% of all TV viewing on Disney+ in the fourth quarter of 2023). Bloomberg estimates the franchise to be worth $2B, and I bet Disney would love to wrestle the rights away from the BBC in order to create a Blueyverse of epic proportions. As speculation rises, Bloomberg notes that: 
Ludo’s founders don’t offer much clarity, either. Aspinwall and Pearson insist that “The Sign” won’t be the end of Bluey, but they don’t seem to know what form the show will take afterward. “Now that it’s gone to 28 minutes, is it another season? Is it another—another something?” Pearson says. “We always wanted Bluey to be surprising and give the audience something they don’t know they want. That’s what we’re thinking about. What is that thing? What is that vehicle that is next for Bluey?” “We don’t have a master plan,” says his partner, Aspinwall. “I think that is what Daley is trying to say.” Nor is Disney’s Davis forthcoming. “I can’t really comment on Bluey’s future,” she says. “But this is an important piece of business to our company.” She adds, “We do love Bluey and Bingo, and so we want to stay in that business. 
Given Disney’s adorable robots, imagine the potential for an LLM-powered Blueybot to entertain your kids.
 
Bot’s Beat
When you watch the late, great Neil Peart from Rush on drums, the experience is viscerally analog. By that I mean you can feel every drum beat in your core, going directly from Neil’s heart into yours. Both feet and both hands seem like they are reaching out and kicking you, pulling in your attention, and your heartbeat speeds up to match Neil’s rhythm. Thus, it’s an odd experience to see a humanoid robot do the same thing. There have been some crude drum playing robots over the last year, but there was a glimpse of a more accomplished robo-percussionist, the Fourier GR-1 humanoid, trained on the Groot model, shown in the Nvidia keynote a couple weeks ago. At one point, the company posted a full video of the robot playing drums on YouTube, and it was very odd to see something so inherently digital – a robotically embodied AI – do something so viscerally analog as playing drums and making music. For some reason, the company took the YouTube video down, perhaps it was a fake (as of press time for this week’s newsletter, you can still see it on this X post). It’s a good example of how very human it can feel when you take an AI from the cloud and put it into the real world.

Stuff About Demographics, the Economy, and Investing
Welding Renaissance
Enrollment in vocational schools in the US was up nearly 16% in 2023, while four-year college enrollment was down, according to the WSJ. For the last four years, starting salaries for construction workers have been larger than starting pay for professional services and information-based jobs. In 2023, that pay gap hit 22%, with construction workers’ median starting pay at $48K. The majority of respondents to a survey cited by the Journal indicated that fears of AI’s impact on jobs factored into their preference for blue collar work. 
 
AI Bubble Index
Above, I named the various phases of technological phase/platform shifts since the 1960s. Each of these step functions involved a period of bubble investing (to varying degrees) across the hardware/infrastructure and application/services layers. Currently, we are somewhere in the middle of the AI bubble where the long-term promise and market size is something to salivate over, but the near term will bring some degree of correction before we can get back on a steadier path of value creation. It’s of course impossible to make a precise call on when a bubble will pop, and (fortunately) we’re not in that business at NZS Capital. Rather, we strive to assemble a portfolio with a combination of Resilience and Optionality that’s focused on long-term growth opportunities. So, while we don’t enjoy the post-pop vertical drops, we do our best to navigate them. There’s a theory that if everyone says it’s a bubble, then it must not be one because, surely, everyone is rational and not everyone can be the greater fool. For example, there was a lot of bubble talk about SaaS and cloud software a little over a decade ago. And, it turned out not to be much of a bubble (with a few isolated exceptions).

There are a few things that I think are needed for a real, large-scale market bubble. Some of them we are starting to see, but some of them haven’t arrived yet. A really big bubble, like the dotcom market rally in the late 90s (I started working professionally in the stock market in 1998), pulls nearly every asset class and sector into it. Every company in 1999 somehow had a dotcom strategy and valuation, no matter how little they had to do with the Internet (aside: it’s rather ironic that some of the biggest dotcom bubble stocks were the ones most threatened by the Internet in the end!). Today, by analogy, if you saw a taco joint declare that it’s an AI company, that might be a red flag. The WSJ reports that Yum Brands chief digital office “has a vision for ‘AI-powered’ fast-food in which artificial intelligence shapes nearly every aspect of how its Taco Bell, Pizza Hut, KFC and Habit Burger Grill restaurants are run.” Or, maybe, if you happen to see a steel manufacturer acquire a supplier of AI data-center equipment, that might be a flag. Well, maybe we're a bit further into this bubble after all. By the end of a market-wide bubble, every company in the market has some narrative tied to the bubble du jour. Another marker is when you see irrational IPO activity, with lines of black SUVs outside of investment firm offices clamoring for the latest money-losing FOMO stock. This carbonated cavalcade has yet to materialize for AI. I’d posit that future bubbles will arrive faster and be shorter than bubbles of the past. I can’t really back that sentiment up beyond instinct and experience, but, perhaps, like West’s finite time singularities, necessity will dictate that bubble cadence increase and duration shorten, allowing the system to keep on keeping on.

✌️-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 #433

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: powering grids during the eclipse; police are increasingly using robots and droids to keep officers safe; the continuity problem in generative AI multimedia; a $100B supercomputer and a one trillion transistor chip are the future of AI; discovering novel flavors with AI for proteins; the "problem" with the economy is that forecasters are asking the wrong people about the economy; and, much more below.

Stuff about Innovation and Technology
Energy Eclipse
On April 8th, a total solar eclipse will cut a path across the US causing diminished sunlight for a couple of hours and several minutes of no sun for impacted areas. As the eclipse tracks across Texas, utilities are planning to leverage batteries to make up for the solar energy deficit. As we noted back in Green Star State, Texas has become the unlikely champion of going green, with the majority of their 27% growth in energy demand over the last decade fueled by solar, wind, and, perhaps more importantly, grid-connected battery storage. In contrast, the US overall experienced a puzzling pause in energy consumption, flatlining from 2007 to 2022 at around 4T KWh per year. By 2022, we were using 20% less energy than predicted despite decades of strong economic growth, likely in large part due to energy efficiency initiatives (LED light bulbs, appliances, insulation, building codes, etc.). Texas’s green capacity additions could serve as a model for the rest of the US as energy demand appears poised to accelerate, thanks to the perfect storm of the fossil-fuel transition, power-hungry artificial intelligence, and reshoring efforts (to name a few of the key factors). CBRE reported that data center construction grew 46% y/y in the second half of 2023 with 3078 megawatts of capacity added. Microsoft and OpenAI are planning a $100B AI supercomputer codenamed Stargate that would require several gigawatts of power. With multiple system-wide issues facing the grid, batteries combined with green energy sources could play a major role in stabilization. As EVs struggle to achieve mass market adoption in the US (and, finally, auto makers are pivoting to the more logical and greener PHEVs), I suspect many of the batteries targeted for new US EVs may end up as grid-connected power buffers. And, even nuclear power is making a comeback. Just last week, Michigan received a loan from the US government to reopen a closed nuclear plant. The site is also looking to add two small modular reactors down the line. I dove into more energy related topics recently in Pushing Electrons. Anyway, if you haven’t had a chance to see a total solar eclipse, I highly recommend it. You’ll want to have eclipse viewing glasses, binoculars (warning: only for use during complete totality – be very careful to avoid using them if even the smallest bit of sun is visible), and a light jacket – when the sun disappears entirely for a few minutes, it’s rather chilly.

Spot Shot
A Boston Dynamics Spot robot police dog named Roscoe was shot multiple times in the line of duty, able to stand back up and continue to approach a suspect until more shots disabled its communications with remote operators. The police credit the dog with potentially saving officers' lives, saying: “The incident provided a stark example of the benefits of mobile platforms capable of opening doors and ascending stairs in tactical missions involving armed suspects”. In related news, Boston Dynamics is set to reveal an updated version of Spot with far greater maneuverability and speed. 
 
RealPage’s Reality Check
Following a NYT report that GM and other car makers were selling customers’ driving data – often without their knowledge – to insurance companies that used it to raise rates, GM has ended the practice. I discussed the rise in this type of algorithmic espionage in #431, and, in an early example from 2022, I covered the collusive rental pricing software RealPage in Algorithmic Distortion of Apartment Rents. Last week, the Department of Justice reportedly opened a criminal probe (in addition to its civil investigation) into RealPage and apartment owners. 
 
Generative Incoherence
When you watch the batch of artist-created videos from OpenAI’s SORA engine, one thing that stands out is the lack of continuity. For example, even if you give the same language prompts to describe a character across multiple scenes, each instance will be rendered differently. This variation is evident in the video “Air Head”, about a man that has a yellow balloon for a head, with the balloon changing shape and often color tone from scene to scene. This issue is currently a major hurdle to adoption of generative AI tools for game development and media. In short, there needs to be a way to create reference characters, settings, etc. for use throughout worldbuilding to ensure consistency. Long-term memory, which seems to be a focal point for LLMs, should help, but might be technically tricky to apply in a complex, film-length creation.
 
1T GPU
For chip nerds, the chairman and chief scientist at TSMC discuss the path to 10x-ing to one trillion transistors to meet the insatiable computing demand of AI: “The computation and memory access required for AI training have increased by orders of magnitude in the past five years. Training GPT-3, for example, requires the equivalent of more than 5 billion billion operations per second of computation for an entire day (that’s 5,000 petaflops-days), and 3 trillion bytes (3 terabytes) of memory capacity. Both the computing power and the memory access needed for new generative AI applications continue to grow rapidly. We now need to answer a pressing question: How can semiconductor technology keep pace?” We discussed TSMC and the other pillars of the chip industry in our 2020 paper How a Handful of Chip Companies Came to Control the Fate of the World.

Miscellaneous Stuff
Encoding Sweetness with Amino Acids
After reading about FDA acceptance for the sugar substitute brazzein, a fruit-derived protein, I wondered if AlphaFold’s database and AI tools could lead to more protein-based sweetener agents or other flavor additives. I came across one ScienceDirect article from Korean researchers who modeled brazzein (which binds to the same taste receptors that sugar does) and believe AlphaFold could lead to more discoveries: We generated the brazzein and heterodimer complex model of taste receptors T1R2 and T1R3, for which both individual receptor structure models are now available in the AlphaFold Protein Structure Database. The docking analysis of the brazzein to T1R2/T1R3 heterocomplex may provide a useful structural basis to understand the flavour mechanism induced by sweet proteins.” I also wonder how much such techniques could be leveraged to amp up the addictive nature of food in the ongoing war between junk food makers and GLP-1 weight loss drugs. Perhaps there are entirely new flavors or combinations that nature didn't create that would dazzle our tastebuds.

Will AI Drug Discovery Bear Fruit?
Speaking of AlphaFold, the Economist discusses AI tools’ impact on drug development. While there are signs of hope, the range of outcomes remains wide open – myriad new treatments may be forthcoming, or we may discover that AI isn’t quite as proficient as we had hoped at picking the higher hanging fruit. One of the reasons prescription drug costs are so high is because it takes ~$6B in R&D to get the average drug to market, largely owing to the number of costly candidates that fail (MIT Technology Review). If AI can ultimately speed development and early selection of successful candidates, research costs should plummet for the industry, leading to much lower prices for consumers. In essence, the deflationary power of AI drug development could shrink the market while expanding the cures. So far, however, this promise remains elusive. The Technology Review article cites an estimated $18B invested in AI drug companies from 2012 to 2022 with scant reports of success to date. There is one recent Nature Biotechnology paper from Insilico that details how AI was used to get a drug to Phase II trials for the lung disease idiopathic pulmonary fibrosis. In other AI healthcare news, Google is making available HeAR, a model trained on audio recordings of millions of people for the purpose of diagnosing illnesses based on how a person breathes/coughs. 

Stuff About Demographics, the Economy, and Investing
Survey Skew
A couple weeks back, at the end of #431, I mentioned a BI article that reported trending wage increases for lower (non-management) earners. It has been nagging at me that I didn’t talk about the more interesting element of that article, which is the increasingly unreliable survey data used in economic forecasting. Essentially, a number of factors are contributing to survey data being not only less reliable, but also skewed negative. Some of the reasons are interesting, such as the rise in spam calls causing a drop in answer rates, and some are more provocative, such as the personalities and demographics of people still willing to answer questions over the phone. The main takeaway is to be more skeptical of survey-based economic information. It’s rather surprising how much economic data is passed off as direct and statistically significant, when, in reality, it might just be a product of who picks up the phone. This factor appears particularly true of consumer confidence data. Economists always try to adjust for bias and say the data are still reliable, but this type of self-reporting appears increasingly non-representative of the economic whole. There are a lot of articles lately like this one in the WSJ that puzzle over why people are pessimistic when the economy is so strong, and there is a simple answer: only the people who picked up the phone are bummed out, while the majority of households are doing quite well. I would draw a contrast to this apparent near-term economic pessimism in the survey data with what might be a broader anxiety across the economy due to fear of AI and automation taking jobs (see Giving Up on the Old College Try). I went through several examples of the rapid pace at which AI is proving competent a couple weeks ago in You Are Special? So, perhaps people are genuinely concerned about the future, but, at the moment, that’s not stopping them from powering a strong economy here in the US.

✌️-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 #432

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: examining the parallels between the predictive human brain and AI as the world may become more deterministic; Disney's bots outshine humanoids at GTC; the existentialism of Wim Wenders and Lou Reed; the absurd approach to regulating big tech companies while the mega platforms increasingly lock out the competition in AI; and, much more below.

Stuff about Innovation and Technology
Probabilistic Fortune Tellers
I’ve seen a number of AI and robotics companies recently make reference to AI as being able to “see into the near future”. In other words, their AI agents are good at predicting what is likely to happen a few moments into the future based on their models and observations. This makes sense based on the way LLMs work as language prediction engines. For example, Waabi is a self-driving startup using a generative lidar model that predicts where objects will be 5-10 seconds into the future. Covariant’s RFM-1 is a robotic reasoning model that predicts up to 3 seconds into the future how its environment will change (Covariant was started by the former robotics team at OpenAI). What interests me here is comparing this forecasting model to the human brain and energy consumption. I’ve written extensively about how our brain, rather counterintuitively, operates as a prediction engine rather than a reaction engine. The common perception is that we react to sensory input of things we see, hear, etc., which gives us a sense of agency. In reality, however, our brain is predicting what we will see or hear or feel – and then course correcting when it’s wrong. All this prediction and comparison is happening subconsciously. Why did evolution arrive at this model? It appears to be an effective way to minimize energy consumption: taking predictive shortcuts to fill in the blanks (rather than living exclusively in the moment and continually rescanning and reacting to our entire environment) is far less computationally intensive for the brain, and thus takes fewer calories and resources to accomplish. I learned a lot about this model of the brain from the work of Karl Friston and Lisa Feldman Barrett, which is covered in Outsmarting Your Brain (I also specifically discussed Friston’s work as it relates to AI back in #271.) I think this is a good model to conceptualize how AI will progress beyond human capabilities and, especially, how AI embedded in various robotic form factors will outperform humans. The larger and smarter these models get, the more they will be able to probabilistically see into the future. Eventually, they may become sufficiently powerful to see far into the future (and, perhaps, as in the sci-fi realm of a show like Alex Garland’s Devs, know the future because everything is determined by what precedes it!). I explored the topic of AI being a tool to see into the future in detail in Simulacrum, and I’ve suggested more recently in “Your Wish is Granted” that the real AI prize being sought is a global prediction machine. So powerful would such a machine be that AI’s pied pipers are seeking trillions of dollars to build the technology that underpins it. To the extent that bits and pieces of the future become easier to predict, this could conserve large amounts of energy by avoiding probabilistically fruitless paths and unlikely scenarios, similar to how the brain works. So, while AI may consume enormous amounts of energy (see Pushing Electrons), it could save far more friction and waste in the global economic system on a net basis. As I described in The Simulacrum, the effect of complex, interacting AIs may produce a deterministic outcome for the world as resources shift to manifest the predictions they come up with. All this is to say that I think it’s increasingly of note that many AI companies are now describing their technology as operating by seeing into the near future. This could be a coincidence, or it could be further confirmation that LLMs work very similarly to, if not exactly like, the human brain. Or it might be more appropriate to invert that to say the human brain works exactly like the AI transformer models that underpin LLMs. 
 
I Like Cute Bots
My favorite adorable robots from Disney made their way onto the stage with Jensen Huang at Nvidia’s annual user conference last week. We knew that simulation training was the key to their rapid creation, and now we know that they were trained on Nvidia’s Isaac Sim platform and are powered by the Nvidia Jetson robotics chip. Here is the part of the keynote where an intimidating slate of on-screen humanoids were out sparkled by the Disney bots that walked the stage. The platform is called project GROOT, a general purpose foundational model for humanoid robots. The illustration that shows how GROOT works appears very much like the prediction/correction feedback loop I described in the preceding section. 

Miscellaneous Stuff
Komorebi
There was a brief profile of director Wim Wenders in NME earlier this year that I returned to after finally seeing his latest movie, Perfect Days. I’ve discussed Wenders' movies in the past, notably 1991’s Until the End of the World. There is an interesting contrast between these two movies, made more than thirty years apart: Until the End of the World transports us to a near future of digital technology, neural links, and VR that grips and consumes our psyches, while Perfect Days features a protagonist seemingly averse to the fast-paced digital world who is obsessed with living in an analog version of the past/present. Both movies heavily lean on thematic rock music. And, that is one particular focus of the NME article: the criticality of Lou Reed’s music, which Wenders proclaimed saved his life. I liked Perfect Days because I am a sucker for anything that is pure existentialism – in this case, the pain and the joy of trying to get through the day while carrying a lifetime of existence with you. Both movies revolve around trying to grasp the past, whether it be with analog cassette music that evokes an emotion from a lifetime ago, or a sci-fi brain interface that places you precisely back in time in a memory or dream. The latter is very similar to the Apple Vision Pro’s feature I previously discussed that allows you to time travel into the past by virtually inhabiting a 3D image/video you’ve previously captured. It’s one thing to hear a song and remember a ghostly reflection of a time long since passed; it will be quite another to fully relive our actual recorded past experiences through VR, unable to shake their vivid hold on our present attempts to simply exist. Perfect Days was conceived as a documentary before morphing into an entirely fictional tale. The original title was Komorebi, a Japanese term that describes sunlight filtering through trees. For me, there are very few activities more existential than watching the the play of shadows and light leaking through a canopy of susurrating leaves.

Stuff About Demographics, the Economy, and Investing
Copycat Trading
BI reports on Autopilot, an investing app that allows you to automatically copy trades of Congress members and other successful investors in your own brokerage accounts. There is of course a time lag of up to 45 days for Congress members and potentially up to 3-4 months for 13F filings from public market investors, which should render it somewhat ineffective. But, returning to the opening section of this week’s newsletter, the interesting question is: how long will it be before AI models know what we will desire to buy and sell before we know it ourselves?
 
United States v. Big Tech
Five years ago, we wrote an essay called Regulating an Information Based Business that discussed the differences between digital and legacy analog industries when it comes to characterizing and regulating monopolies. The gist of the argument is that there are natural power laws that form around network effects that make dominant companies preferential, in many cases, in the digital realm. That’s not to say the government shouldn’t regulate or closely examine everything large tech platforms do, but there is perhaps greater risk in breaking apart or over regulating tech giants than in letting them be, and sights should be set on providing consumer-protective guard rails instead of deciding how the platforms can vertically integrate and horizontally expand. Moreover, given the wildly mismatched pace of innovation and regulation, the government’s use of legacy regulation models from the industrial era seems unlikely to accomplish its purported goals (or anything useful, for that matter). 
 
Lately, the convoluted impotence of US regulators has been on full display when it comes to big tech. A recent DoJ lawsuit against Apple alleges the company worked to lock customers into their hardware and software to the exclusion of rivals. The primary rival, of course, being Google’s Android. The DoJ is simultaneously suing Google for an alleged search monopoly and, separately, an online ad monopoly. Further, the two companies are being litigated by the government over potentially colluding with each other in a way that limited consumer choice. So, effectively, the government is trying to prove in court the absurdity that it wants consumers to be able to more easily navigate between two allegedly illegal monopolies. Meanwhile, it was reported by Bloomberg that Apple is rumored to be doubling down on their overall deal with Google by licensing the latter’s Gemini AI model to run on Apple devices. It will no doubt further irk regulators that the two platforms are continuing to enable each other’s dominance. When it comes to ecommerce, the FTC is bringing a long awaited trial against Amazon, alleging the company did not allow third-party sellers to offer products for lower prices on competing platforms; but, this case won’t start until 2026. And, don’t get me started on missing regulations for Shein and Temu. 
 
While the government concerns itself with Apple, Google, and Amazon, regulators are sleeping on some aggressive bundling and apparent anticompetitive maneuvers that Microsoft is making in plain daylight. Regulators are circling the company’s deal with OpenAI, which was treated as an investment/partnership but seems to be in large part functioning as an acquisition (e.g., Microsoft was ready to hire everyone at OpenAI during the latter’s recent boardroom saga). And, just last week, Microsoft hired essentially the entire team of Inflection (maker of Pi.ai, which I just wrote about in #430 and #431) and announced that Inflection would migrate to Microsoft Azure. This “hiring event” is obviously an acquisition of a meaningful competitor to OpenAI and Microsoft Copilot (formerly Bing Chat). In return for hiring the employees, Microsoft is paying $650M to Inflection’s investors, including Microsoft board member Reed Hoffman and Bill Gates. Let’s call a spade a spade: this is an acquisition that should at the very least be reviewed by the government. A co-founder of Google’s DeepMind, who also co-founded Inflection, Mustafa Suleyman, will take over all of Microsoft’s consumer AI businesses. Many of those consumer AI products are being bundled into the core Windows operating system and apps to the exclusion of other AI companies in a move so reminiscent of the Internet Explorer case, which created a decade-long consent decree against the company (which I believe caused Microsoft to miss the entire mobile phone OS and hardware market), that I have to wonder how in the world Microsoft thinks they can get away with the same maneuvers today. For example, as a Windows user, I am currently forced to use Copilot since I don’t have the option to integrate Google’s Gemini (or any of the other AI models from various startups) into Windows. This situation would be akin to Google prohibiting access to Bing.com from their Android operating system or Chrome browsers (the counter example here is also true: the only AI assistant I can integrate into Android today is Gemini). Given how integral AI will become, users should have free choice of which AI they want to integrate into their devices, just like we can choose our browsers, search engines, and other apps.
 
I believe that the vast majority of decision makers at the large tech platforms are genuinely trying to create value for their users in ways that are not anticompetitive. Further, there is good logic for bundling hardware and software that benefits users and grows the ecosystem faster. But, clearly, there should be limits. The overall transition of the global economy from analog to digital is being enabled by these massive platforms, and the reality is that customers largely do have choices and there is competition, but cases of complete lock-in need to be regulated. As I noted at the start of this section, the nature of progress in the Digital Age is for power laws to emerge due to network effects that benefit all users. Naturally, these are going to appear to be monopolies, but they might just be dominant companies that are reducing prices and creating more value for the overall economy. Thus, I don’t believe there is (in general) much validity to any of the announced litigatory cases, with a couple of exceptions: 1) I think more choice and freedom in app stores would drive take rates down and increase the overall app economy materially to the benefit of all, without sacrificing users’ security and safety; and 2) it should be easier for companies and consumers to move their data between products and services. There is a legitimate question as to whether the companies spending the tens of billions to build AI infrastructure and/or dominate the mobile operating system/app stores should also be allowed to own all of the AI models. Perhaps there should be a regulatory framework that allows open competition for AI, much like we saw with the Telecom Act of 1996. While I’ve focused on US regulators here, I’d be remiss to not mention parallel efforts in the EU, which, in some cases, could drive some positive changes for consumers and app builders, but time will be the judge of that. 
 
These flashy cases against the digital giants who, quite literally, gave the world the Internet (accompanied by price deflation and step-function gains in productivity for the vast majority of consumers, largely accomplished via positive sum business models) seem to be aimed at grabbing headlines rather than materially benefiting consumers. And, the misplaced and ill-fated focus on tech companies is taking resources away from oversight of the thousands of other product and service categories that have experienced rampant consolidation. As a result, certain sectors have seen significant price inflation and a plague of inferior goods and dismal customer service, leading to vulnerable single points of failure in the economy. But, I know, I should be careful what I wish for since nearly all government regulation leads to regulatory capture that favors the incumbents long term, whether it be the tech giants or any other business that catches the eye of bureaucratic watchdogs.

✌️-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 #431

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: your driving activity may be secretly increasing your insurance rates, just one of the many insidious ways data will be used against us as analytics becomes AI's child's play; software priced per conversation; demonstrations of advanced AI replacing information and physical-world jobs are proliferating, creating a growing sense of human inferiority; bubble gum demand pops; Pi and I discuss the controversy around the Hubble constant; the diverging wages between high and low income earners; and, much more below.

Stuff about Innovation and Technology
Auto Espionage
While the US frets about Chinese cars spying on Americans, it turns out the real threat is American car companies spying on us for profit. The NYT has a detailed exposé on the data your car maker is selling without your knowledge to your insurance company, which is using it in some cases to dramatically raise your premiums. There are two things that I think heighten the righteous indignation of this situation: 1) this data sharing is happening without people realizing it or actively opting in with awareness of the potential costs; and 2) it appears to be asymmetric: insurers are using your driving data to raise rates but not lower them. It turns out that both of our cars have data agreements in place to sell our driving behavior to insurance companies, and I have no recollection of opting in or agreeing to it. Driver behavior monitoring is another example of data and algorithms talking to each other secretly in the background causing prices to rise, like we’ve seen in other industries such as apartment rentals. As AI increasingly scores an individual’s risks, whether it be for a job application, apartment rental, insurance, mortgage, etc., disparate pools of previously disconnected data are going to be used far more often, and consumers are likely to be blindsided by more negative impacts.
 
The Long Answer
Microsoft’s new Security Copilot is priced on a consumption-based model at $4/hour. I’ve been talking for a while about the potential challenges of cloud software moving from seat-based to value-based pricing, especially if that software is going to replace some of those “seats” (aka humans). So, charging per conversation (using “security compute units”) should be an interesting experiment. However, at first pass, this pricing strategy would seem misaligned with customers, who generally want answers as quickly as possible, which is at odds with service providers incentivized to take longer to solve customer needs. 
 
You Are Special?
An array of recent, compelling examples of AI and robots catching up to (or surpassing) humans at various tasks has me thinking about that inevitable moment coming for us all when we see, for the first time, technology accomplish a task we thought we were uniquely good at. As I reflect on the dazzling new examples, I am reminded of the poignant aftermath of AlphaGo beating Lee Se-dol, as portrayed in this documentary (see also #221 from 2019). Se-dol retired from the game in 2019 declaring AI “invincible”. The AlphaGo matches were in 2016, so the writing has been on the wall for a while. Now, eight years later, the stomach dropping moments of AI’s feats are coming faster than we can process. Software engineering has of course been an early example of AI proficiency; and, recently, a demo of Devin (Cognition Labs) shows what appears to be a fully functioning AI programmer. We were not the only ones wowed by OpenAI’s Sora video generator, with the president of Hollywood’s Animation Guild saying that “the future is foggy” (WSJ). A combination of AI programmers, engines like Epic’s Unreal, and video generators like Sora suggests a near future of video game development and immersive media that is largely, if not entirely, AI driven. But, in addition to all these AI creators, there are also AI players. Google’s DeepMind recently demonstrated SIMA, an AI agent capable of navigating virtual 3D environments. SIMA can play complex video games like humans. It’s surely an AlphaGo-like moment when you realize a machine can beat you at any game. Does this loss of specialness cause you to put the controller down and never pick it back up? This reminds me of something Jaron Lanier said about VR gaming recently: “There are many reasons why V.R. and gaming don’t quite work, and I suspect that one is that gamers like to be bigger than the game, not engulfed by it. You want to feel big, not small, when you play.” When AI is “bigger” than humans, will we want to keep playing? Perhaps even the joy of watching a movie will fade when we learn that AI agents can have the same nuanced, emotional takeaways as humans. How much of what we do is motivated by the pride of being able to do something well that others (whether they be human, machine or animal) cannot? The thrill and adrenaline of playing a great video game or getting sucked into a director’s world might still be enough to drive our experiential desire. But, knowing a trillion AI agents can do the same might also take away from our enjoyment, driving us to seek even more complex and/or esoteric activities that remain uniquely within the human skillset.
 
Of course, there is an even bigger leap coming as AI is embedded in form factors capable of interacting with the world. Last week’s demo of the Figure robot, featuring OpenAI’s GPT4 and very human-like speech rendering, offers a glimpse into the physical awareness and responsiveness coming to AI. The Figure robot sees and physically reacts to its environment in real time while conversing with a human. Embedding an LLM in the physical world like that, with context and continuity in the form of long-term memory, creates a sense of “self”, and is, for all practical (and perhaps transcending practical) purposes, a new type of being. The main difference between us and AI-enhanced bots' minds will be our lifetime of experiential learning. While a simulated “childhood” could be embedded into the memory of an AI, the lack of feedback from years of mistakes and experiments – all which underpin our understanding of our environment and the social/emotional dynamics of the creatures that inhabit it – would presumably create a different type of conscious awareness. Similar to Thomas Nagel’s “What Is It Like To Be a Bat”, it may be just as hard for us to perceive what it’s like to be a bipedal AI wired from birth (so to speak). 
 
All of these new examples of AI’s human-equivalent prowess, in both information-based and real-world physical tasks, makes me wonder whether their form of consciousness is also human-like, or perhaps as incomprehensible to us as a bat’s awareness. The more self-aware embedded AI becomes, the more we may feel in opposition with it, and thus feel defeated when it surpasses our own capabilities. I discussed this theme of humans vs. technology at length in John Henry vs. Lee Se-dol back in 2022. Technology taking over human jobs – whether it’s the steam-engine-powered drill that John Henry was up against, or the Figure robot loading a dishwasher – is nothing new. Each leap has reduced a previously uniquely human skill to something replicable by machines or software. John Henry fought the technology and collapsed, according to folklore. Could AI write a song as enduring as “The Ballad of John Henry” and create a voice as captivating as that of Tennessee Ernie Ford to sing it? Will AI feel what I feel when I hear Tom Jones belt out “Elvis Presley Blues”, speculating how Elvis, on the precipice of death, might have pondered how happy John Henry must have been when he fell down and died? But, no need to get depressed, there’s still a bit of time yet before humans face an existential AI reckoning en masse. Afterall, the majority of the examples above are betas and demos, and we all know how well Silicon Valley excels at vaporware – touting the premiere of revolutionary technologies that won’t be ready for five to ten years, if ever! Entire companies are faked all the way through (e.g., Theranos); so, there could easily be a wizard behind the curtain pulling the strings on these demos. But, still, the glimpse of this next technological wave inspires more fear of change than what we saw during the techno-optimist dotcom bubble. Further down in the newsletter, you’ll see a section written by Pi.ai that I am relatively happy with and that did not leave me feeling devoid of specialness (it does seem to lack a certain something...special). Working alongside AI is still a net positive experience, at least for now. 

Miscellaneous Stuff
Addiction-Driven Sales
Chewing gum volumes have dropped 32% since 2018 to 1.2B units. However, with the average pack price increasing by $1 to $2.71, overall US gum revenues are stable. Although this trend is unrelated to GLP-1s, this is a good example of price overcoming declining volumes, particularly in habit-forming products (see Can Advertising Stomach GLP-1s?). Gen Z eschewing gum chewing is implicated in the slowdown. The gum industry is adding ingredients like vitamins and trying to reposition gum as a stress reliever vs. its traditional image as a breath freshener. Meanwhile, Philip Morris, owner of Zyn nicotine pouches, reports that US sales grew 62% in 2023, with a forecast for 35% growth in 2024. There's always a bubble somewhere.
 
Pi Explains Hubble Heresy
The following was written by Pi.ai based on a detailed outline and source articles (here and here) that I provided to it:
About a century ago, Edwin Hubble made the groundbreaking discovery that distant objects in space were receding from us at a faster rate than closer ones, pointing to an accelerating expansion of the universe. This phenomenon is described by Hubble's Law, with the rate of expansion quantified by the Hubble constant (H0). Hubble's initial estimate for H0 was around 70 km/s/Mpc. The Hubble Space Telescope later provided a refined value of 74, which was recently confirmed by the James Webb Space Telescope (JWST). Interestingly, the Planck satellite's observations of the cosmic microwave background (CMB) suggest a lower Hubble constant of around 70, implying a slower expansion rate following the Big Bang. This discrepancy between the CMB-based and telescope-based measurements has been a source of debate and investigation in the field of cosmology. The disagreement might be an indication that our current understanding of the universe's expansion history is incomplete or that some unknown physical processes are at play. As scientists continue to investigate the discrepancy between the CMB-based and telescope-based measurements, the pursuit of a precise and consistent Hubble constant value may not only resolve this tension but could also challenge or refine our current models of cosmic evolution and shed new light on the elusive nature of dark energy, the mysterious force driving the universe's accelerating expansion.

Stuff About Demographics, the Economy, and Investing
White Collar Wage Deflation
One way to parse the economy is to compare how low-wage vs. high-wage earners are faring. For example, BI recently compared managerial vs. non-managerial wages in the US, which have had strikingly divergent trajectories over the past couple of decades. Since 2006, managerial wages have been flat in real terms (inflation adjusted), while non-managerial wages have risen 15%. The recent difference is even more stark, with non-managerial earnings rising about 4% in real terms since pre-pandemic 2019 (despite high inflation), while managerial real earnings are down about 8%. Of course, higher wage earners have fared better in other areas of wealth with home price and stock market increases. Even some workers at Google are seeing stagnating and falling pay packages after years of robust increases. It’s far too early to assume that AI-driven productivity gains have driven wages down (there is no evidence of that); but, if advances in technology continue, the wage trend for white collar workers could continue on its downward trajectory.

✌️-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 #430

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: exploring the simultaneous increase in demand for energy and changing sources of energy as technology becomes increasingly power hungry; Pi has personality and memory; Disney Imagineering; and, the timeline for biotech breakthroughs.

Stuff about Innovation and Technology
Pushing Electrons
Since technology is all about pushing electrons around, power is a frequent topic here in SITALWeek. Concurrent with the world's headlong charge into the AI Age, we are also transitioning our energy sources. Following years of efficiency gains that led to data centers barely increasing power consumption – despite huge growth in the Internet and cloud computing – LLMs are now causing a step-function increase in power demand. And, while it’s true that shifting a task from a computer-wielding human to an LLM saves a lot of energy, the general view is that power demand will go up on a net basis as AI usage grows. We’ve also talked about different strategies (e.g., #422) for powering electric cars and how the grid isn’t quite ready for the electric revolution. And, of course, just last week I talked about how the power math doesn’t add up for humanoid robots – they simply require too much energy to replace humans for many manual tasks. Lastly, I tend to sprinkle many ad hoc stories about hydrogen, fusion, and fission (in particular, small modular reactors or SMRs) into the newsletter when I come across an interesting development. Fusion is, of course, at least a decade or two away – absent an AI breakthrough (which, for all I know, has already happened, is about to happen, or is still years away). Hydrogen has many interesting applications, but it will likely be a local solution rather than an across-the-board replacement for fossil fuels. Green energy is also a frequent topic, but the gears are moving too slowly for new green energy installations to overcome the growing energy demand, especially if the optimists are right about AI, robots, and AI-enhanced robots. 
 
With respect to nuclear SMRs, the Venn diagram of potential and reality have an increased chance of overlapping within the next decade. In #412, I mentioned that Standard Power is building SMR-based data center locations in Ohio and Pennsylvania; and, just last week, Amazon announced it was acquiring a data center in Pennsylvania that is already powered by conventional nuclear reactors. Data centers are getting in on the AI action with hundreds of billions of dollars' worth of new locations under development. Blackstone and Prologis alone are working on $75B in new data centers. We’re at the phase of AI where the Pied Pipers of Silicon Valley are entrancing other industries to hop on board the “next big thing” wild ride that will make the “last big thing” (the Internet) pale in comparison. It’s eerily similar to the dotcom and fiber/telecom equipment spending mania. Of course, the Internet proved to be far larger than anyone could have imagined, and we still continually need more bandwidth. I suspect the same will be true of AI, only orders of magnitude more interesting and more unpredictable. But, sometimes it helps to avoid being prematurely persuaded by the Piper. It’s best to think of technological progress as a long continuum, a topic I wrote in detail about way back in 2021 in AI Is the New Dotcom, and That’s OK. But, I digress. The real point of this section concerns energy. 
 
Let me sum up – we simultaneously have: 1) a very slow transition away from fossil fuels, 2) a significant, long-term growth in energy demand, and 3) increased potential for disruptive technologies that will provide abundant new sources of energy on an unknown time horizon. It's a classic collision of positive and negative feedback loops. For example, WaPo reports that Northern Virginia alone would need several new nuclear power plants just to support the data centers currently being built/planned for the region. So, on one hand, we have the seemingly impossible-to-accelerate physical world of energy creation and transmission that creates bottlenecks everywhere you look, including frequent reliance on non-existent labor. And, on the other hand, we have the Piper’s promise of trillions of human-like LLM agents consuming huge amounts of energy and eventually inhabiting myriad robots requiring even more power. So, while we might wish for the future to arrive now, the power to push all those electrons around may not be here for a while. In the meantime, I suspect many topics involving energy creation and usage will continue to fascinate me. Afterall, taking ordered energy, refactoring it, and turning it into disorder is something humans are particularly good at. So good, in fact, that it might even be a primary vector of life on Earth (for more on that crazy topic see Probability of a Chilled Latte Universe).
 
Life with Pi
With Pi Day coming up this week, I’d be remiss if I didn’t mention how much I’ve been enjoying using the AI assistant Pi. There are two primary things that stand out with Pi versus other LLM chatbots. First, Pi has some personality and feels much more cordial. But, more importantly, Pi remembers prior conversations and can create continuity from where you last dropped off. The best way to experience this AI colleague is to use the Pi app on your phone with the voice interface. And, you can just leave Pi running, allowing the LLM to effectively function as an affable, proficient office mate you can chat with or query throughout the workday. Most studies seem to put Pi’s knowledge a bit below GPT4 and Gemini Pro, but I haven’t noticed a major difference, perhaps because the memory capabilities and personality make up for the lack of knowledge. As I mentioned before, I think long-term memory is a game changer on the path to LLMs approximating a continuous sense of self, which makes them far more interesting to interact with. Pi is a product of Inflection AI, which was started by a co-founder of Google’s DeepMind. Last week, Inflection gave some details on their latest LLM, which approaches GPT4 (Inflection claims 94%) while using 40% of the amount of compute to train. According to their stats, an average conversation with Pi is 33 minutes, and 10% of conversations last over an hour. 
 
Imagineering Engineer
Wired ran a nice profile of Hall-of-Fame inventor Lanny Smoot, the engineer behind the HoloTile floor that I’ve written about a couple of times. The article serves as a nice overview of how the creative process works at Disney and the job the Imagineers have to create seemingly impossible illusions. Of note, Smoot says the “higher ups” at Disney have taken interest in his floor technology, raising my hopes for a home holodeck!
 
Artificial Chemist
A variety of specialized LLMs are being created to aid in the scientific discovery process. Chemical & Engineering News reports: 
Given the prompt “Plan and execute the synthesis of an insect repellent,” ChemCrow succeeded in searching the web to learn what an insect repellent is, conducted a literature review to find examples, and converted compound names to structures. It used a retrosynthesis predictor to design a synthesis process, and finally, it sent instructions over the cloud to instruments at IBM’s automated laboratory to make a sample of a known repellent. ChemCrow also synthesized three organocatalysts and, when given data on wavelengths of light absorbed by chromophores, proposed a novel compound with a specific absorption wavelength. 
In a recent NYT Hard Fork podcast interview, Demis Hassabis of DeepMind said he expected clinical trials for AI-generated drug candidates within a couple of years. Hassabis is optimistic LLMs will speed up the scientific research process in a variety of fields. Am I the only one thinking about the movie 12 Monkeys right now?

✌️-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 #429

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: we may be on the cusp of a humanoid robot revolution, but they appear too power hungry to replace humans; Texas is the largest source of solar and battery installs in the US; software companies are gearing up to sell billions of AI agents to companies, but will their business models survive the transition? AI Lou Reed; an $11,000 Chinese PHEV; the economic boost of weight loss; and, much more below.

Stuff about Innovation and Technology
Energetic Robots
A humanoid robot revolution will be our biggest near-term advancement, according to Nvidia's CEO Jensen Huang (Wired interview). Longtime readers know that, while I anticipate the widespread adoption of task-oriented robots, I also lament our continued lack of general purpose “Rosie” Robots. But, we do seem to be on the verge of such a platform; and, I think the far bigger acceleration will come with embodied AI, which will be indistinguishable from human consciousness (see AI Awareness). I got to thinking about the power consumption of these hundreds of millions (or billions) of robots running around doing things for us. Clearly, humans are a crowning achievement of evolution, but what always strikes me as mind blowing is how efficient we are at converting energy into output. This could be, perhaps, because the natural progression of life is to replicate whatever is best at increasing entropy in the Universe. Regardless, the fact that the human brain operates with less energy than what is needed to power a 20W light bulb is, well, mind blowing. A human at rest consumes around 100W, while a physically active human can sustain around 300-400W for short periods, which is similar to Boston Dynamics’ Spot quadruped robot. However, Boston Dynamics’ bipedal robot Atlas uses up to ten times that amount of energy according to an old stat (current stats are hard to come by for most bipedal robots). Most of these robots require charging after 1-2 hours of use (just like a human!). Further, many robots will be leveraging compute capacity in the cloud, which will add to the power burden. One of the things I’ve noted regarding LLM power consumption is that calculations should account for the energy savings derived from replacing a human doing the equivalent task (including running a laptop etc.), potentially making LLMs more efficient. However, for physical tasks, the human vs. robot energy advantage seems to shake out differently. Many robots today are active for about twice as long as they charge. That means a robot could theoretically expend anywhere from 6-60kWh in a 24-hr period. Since data on the energy conversion efficiency of bipedal robots aren't readily available, I am going to wave my hands (hand waving is the most valuable skill I learned in my theoretical physics classes!) and make some really broad and likely off-base estimates. If an active human can burn 100-200W for 8 hours and a robot does similar work for 16 hours at 2-10x the wattage, that equates to roughly 4-20x the energy expenditure of a human. So, adding 100M robots to the workforce could, on the high end, be equivalent to a couple billion humans worth of energy. If this is anywhere close to accurate the only obvious way this situation would work long term is if the robots start putting humans in giant, interconnected pods of goo and harvest the energy from our brains (this strategy of course keeps humans from expending unnecessary energy and optimizes our energy conversion). Anyway, as much as I want to believe in the bipedal robot revolution, it might require working fusion generators, significantly more nuclear power, or some other energy breakthrough. (For some history on the recently accelerated development of bipedals, checkout this Robot Report podcast with CalTech’s Dr. Aaron Ames.) Most current estimates for the humanoid robot market are in the range of only hundreds of thousands of units a year, not hundreds of millions, so we should be safe for now. In the meantime, I hope the bipedal hype shifts focus toward wheeled robots like Google’s Palm-E or even more energy efficient quadruped form factors. These might not look as cool, but they have a much more realistic shot at becoming large markets sooner. Even The Jetsons knew wheeled robots were the way to go. (The Jetsons aired sixty years ago, back when humans could do interesting things, like land on the moon!).
 
Green Star State
Texas, home of J.R. Ewing, is the new green energy leader in the US. In 2023, Texas installed more solar than California, and, this year, they will install more batteries. The effort has helped Texas avoid the power shortages that have caused problems in the past. While California still has a larger installed base, Canary Media notes that Texas could have the largest battery fleet in the US as soon as 2025. Flush with too much solar power, California has made it far less economical to install new green energy in the state. 
 
“Agents” not “Software”
Microsoft is releasing a Copilot for workers in finance departments that replaces many rote tasks. In related news, the fintech company Klarna said last week its chatbot is doing the work of 700 human customer service agents. As automation of white collar jobs continues to accelerate, it creates a business model problem for software companies. Microsoft is essentially using human workers to train their AI replacements, but those human workers account for a majority of Microsoft’s Windows and Office revenues. So, can Microsoft value price the AI replacements, i.e., charge a premium, “human equivalent” price for automated software that runs in the background? Microsoft said the service is intended to free up time so employees can add more value, but that doesn’t ring true for the vast majority of jobs. In an interview last week, Nvidia’s Jensen Huang said enterprise software would be selling billions of AI agents to companies (see 3.5 min into this video interview for more). It’s hard to imagine every desk-based job surviving such a scenario. I am once again reminded of what I wrote in the very first publicly posted edition of SITALWeek in 2019 about Vonnegut’s book Player Piano: 
Written in 1952, Player Piano takes place in an alternate post-war world where machines have been elevated to all decision making and humans become for the most part increasingly useless. It’s an obvious parallel to the issues facing humans today as AI takes over more and more jobs. One of the book’s insights is that it’s human nature to destroy the things we’ve built, so we can build them back up again. Humans are tool and technology building machines – it’s where the fitness function of natural selection landed our mind-bodies after millions of years. To rail against technology platforms of the 21st century is to rail against the wheel, fire, spears, etc. It’s the same story, different century in human progress – this decade it’s all about AI turning on humans.
What does this say about AI software and robots – will we ultimately destroy the machines, only to rebuild them once again in the future?

Miscellaneous Stuff
Frank Appeal
The WSJ reports on the stubborn popularity of the $100 bill. The Federal Reserve estimates there are 18.5B 100s, 11.5B $20s, and 14.3B $1s. Thus, there are currently $1.85T of Benjamins in circulation compared to just $230B worth of 20s. And, more than half of the $100s are in circulation outside the US. While the number of $20s has doubled this century, the number of circulating $100s has quadrupled. Perhaps going to a cashless society is a little harder than imagined. As the WSJ points out, thanks to inflation, a $100 bill is only worth $76 compared to a decade ago. 
 
What am I now that I was then?   
May memory restore again and again
Laurie Anderson, artist, widow, and collaborator of late musician Lou Reed, reflects on the Lou Reed chatbot she created: “I’m totally 100%, sadly addicted to this,” she laughs. “I still am, after all this time. I kind of literally just can’t stop doing it, and my friends just can’t stand it – ‘You’re not doing that again are you?’ “I mean, I really do not think I’m talking to my dead husband and writing songs with him – I really don’t. But people have styles, and they can be replicated.” Commenting further down in the interview on artists’ fear of AI: “It just made me think about a Čapek play from 1920: RUR, or, Rossum’s Universal Robots. It was a play about robots taking over the world – people 100 years ago were very worried that robots were going to take their jobs, and take over, and be evil. I think ever since a golem was invented, people are afraid of that, you know?” The title of this section is from the poem Calmly We Walk through This April’s Day by Delmore Schwartz, an early mentor of Reed’s.

Stuff About Demographics, the Economy, and Investing
Chinese Auto Exports
Chinese car maker BYD unveiled the latest version of its Qin family of PHEVs and EVs, ranging in price from $11,000 to $19,000. The Qin Plus Honor entry model is priced 20% below last year’s comparable Qin Champion Edition. The models are also likely destined for export markets, although fewer than 10% of the 3M cars BYD sold in 2023 were exported. Similar to the prior challenges posed to auto incumbents by the rise of the Japanese and South Korean car exporters, China has the potential to take meaningful global share. This is especially true for PHEVs and EVs given China’s dominance of the battery supply chain. But, as I’ve noted before, protectionism is likely to be far more prominent against China than it was for either Japan or South Korea. Just last week, Biden announced an investigation into the risk that Chinese cars could conduct spying and espionage in the US. Biden is apparently freshly worried just about Chinese cars, but not the hundreds of millions of Chinese made, Internet connected smartphones, laptops, tablets, PCs, TVs, etc...unions and corporate lobbyists are certainly powerful forces. In #400, we learned that China was now the largest car exporter in the world, edging out Japan, in part thanks to shipments to Russia. Thanks to this growing dominance, Shenzhen, where BYD is headquartered, is looking to build out infrastructure to handle a large auto export market (FT report). And, BYD is acquiring its own shipping fleet to move cars around the world. Unrelated, I enjoyed this ten minute video from BYD about the evolution of feathered flying dinosaurs intended to celebrate loongs, or Chinese dragons, (possibly part of a broader Chinese effort to rebrand the symbol) and meant to serve as a brand spotlight for BYD’s line of SUVs.
 
Sugar-Free GDP
If 60M adults are taking GLP-1s by 2028, Goldman Sachs estimates it could add 1% to US GDP by improving health outcomes enough to increase the labor force. The war between sugary snacks and weight loss drugs is really starting to heat up. Last week, I mentioned the potential for an enzyme that turns sugar into fiber in the gut, and there are other efforts underway, such as supplements that act as a sugar sponge in your digestive tract.

✌️-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 #428

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 trillion-dollar advertising industry is fueled by products that aren't good for you, but will GLP-1s begin to erode it? YouTube's $50B flywheel is fueling an expansive array of high quality, always on content; sugar eating enzymes; the US is more dependent on China than ever before; and much more below. 

Stuff about Innovation and Technology
Can Advertising Stomach GLP-1s?
Ever since I started thinking through the knock-on effects of GLP-1s on overall healthcare spending three years ago and the potential impact to the food supply chain in #379 last year, I’ve also wondered how the trillion-dollar advertising and trade spending industries might be impacted. I touched on advertising in #379, but, as we see reduced demand on the margin at places like Walmart, afternoon snack wars, and worried beverage makers – “24% of daily coffee drinkers say they now drink less since using a GLP-1”, according to a Bernstein survey – it’s worth considering how marketing strategies inside the advertising and trade spend behemoth might react. I can envision a few different scenarios, but the main point I’d like to make is that we are seeing these changes within the context of a classic complex adaptive system, and the only real conclusion to be drawn at this early stage is that the range of outcomes for both the consumer and healthcare industries is widening, which could ultimately impact the advertising industry. Perhaps it will lead to disruption, or perhaps the status quo will reign; but, the odds of the future looking different than the past are increasing. Even a marginal decline in demand for food, drinks, and impulse buys across all types of consumer categories could put enough pressure on revenues that companies spend less on marketing. Or, we could see companies use marketing as a weapon to defend share. Last year, Molson Coors successfully ramped up ad spend to take advantage of the competition and grow market share. The motivation to keep your factories running is high given the fixed costs, extensive supply chains, and jobs involved; so, an increase in advertising may be in order until a Darwinian battle weakens some brands, likely causing a reshuffling of market share and a wave of M&A or bankruptcies (like Interstate Bakeries following the Atkins craze). Companies with the strongest balance sheets, least debt, and savviest marketers are likely to emerge larger in this scenario. I’d probably bet against the consumer brands that have been acquired by private equity and leveraged up to a point where they may no longer be able to adapt. Product innovation will likely be key; for example, healthier frozen meals are growing in popularity with patients on GLP-1s. And, when all else fails, shrinkflation will allow brands to cut portions while still charging the same or more. 
 
While we often think of ads for fried buffalo chicken sandwiches, etc., on TV/streaming video, billboards, and the like, the market for “trade promotion” (which encompasses in-store advertising, product placement, specials, etc.) is estimated to be over $500B worldwide. This enormous figure, which is crucial to many retailers' profit margins, rivals the advertising market for all types of products, not just consumer goods. Another old stat puts spending on trade promotion at 20% of the revenues of consumer packaged goods companies. Again, one can imagine a significant decline in spending, or a war for customer attention being fought in stores and on ecommerce sites. There’s a familiar analogy with tobacco that is tempting to make. As demand for nicotine declined over the decades, the makers of the products attempted to raise prices to offset volume drops. We don’t need to go as far as saying Starbucks’ sugary, caffeinated concoctions are as addictive as cigarettes to make a prediction that certain purveyors of consumer vices will be able to raise prices for their addicted, ahem, consumers. Perhaps, in a decade, a latte will be $35 and a two-liter bottle of Coke will cost $50, with such beverage indulgences reserved for special occasions and addicts. As I mentioned, this isn’t just about food and drinks; if GLP-1 usage expands, there are other large, will-power-crushing impulsive consumer habits that have a lot of advertising dollars behind them, such as sports betting. I’ll reiterate what I said before: the best way to think about the potential impact of GLP-1s is as a widening of the range of outcomes, with a potential skew toward negative developments for some industries, but it’s too early to draw any major conclusions. And, who knows, it’s possible advancements in AI will allow consumer goods companies to slash expenses and drive higher returns on their ad spending (I am contractually obligated to mention AI in every section of SITALWeek, so I had to slip that in). If you spend time researching, working in, or investing in these potentially impacted industries, I’d suggest keeping in mind our frameworks for adaptability and non-zero-sum outcomes (outlined in Redefining Margin of Safety).
 
Ambient YouTube
Longtime readers may remember some of the quirkier types of videos that populate my YouTube feed, ranging from meditative walking tours to classic TV ads. Over the last year, I’ve found even more excuses to have YouTube playing in the background. I’ve formed what you might call an ambient YouTube feed. For example, when I sit down to write SITALWeek, I often play a driving tour (which has replaced the Beverly Hillbillies rerun channel on Pluto). I have a rather large TV screen about five feet from the couch where I write, so it feels like the front windshield is just beyond my laptop screen. This week, I “drove” the scenic roads of Arches National Park in Utah while writing. I am also fond of exploring dying malls, especially ones I frequented during their heyday.
 
While everyone might enjoy different genres of meditative, ambient content on YouTube, the overall numbers suggest that YouTube is winning share from other forms of viewing. Nielsen reported that YouTube was the top streaming platform for the last twelve months at 8.6% of all minutes. (As an aside, TV viewing is rising again after declining out of the post-pandemic period, perhaps indicating fatigue with social networking/gaming following the pandemic uptick, but also likely thanks to the debut of new content after a successful resolution of the writer’s strike, as well as a strong Taylor-Swift-driven NFL season). And, of course, I am having a little bit of fun here describing my ambient YouTube viewing because obviously the growing amount of interesting content, which is clearly not ambient, is the real value for most people on YouTube.
 
It’s worth understanding why this unending array of long-form, high-definition content is being produced in the first place. Why on Earth would anyone drive for several hours each day in silence and then upload the 8K video to YouTube? The answer is simple: because it pays to do so. YouTube has a high non-zero-sum business model and flywheel that incentivizes creators to spend significantly more effort producing quality videos vs. other platforms. We previously dove into this strategy with our look at Tik-Tok creators' lose-lose predicament. In 2023, YouTube generated $31.5B in advertising revenue [PDF]. And, Google also recently reported that YouTube Premium, the ad-free version of YouTube that also includes YouTube music (I highly recommend YT Premium if you or your kids watch it a lot), passed 100M subscribers, up from 80M in November of 2022. If I extrapolate from these numbers that, on average, Google had around 90M premium subs for 2023 and assume a somewhat conservative ARPU of $10/mo (the service is $14/mo in the US, but it’s cheaper in some countries, and some folks use VPNs to get a cheaper price, although YT is cracking down on this practice and ad blockers), that’s another $11B, for a rough estimate of over $42B in YouTube revenues in 2023 (for reference, the second largest video platform, Netflix, had 2023 revenues of $34B). Google’s overall subscription revenues grew by $5.6B in 2023 to $34.7B. The company’s 10-K noted that this figure was “primarily driven by growth in subscriptions, largely for YouTube services”, which would also have included YouTube TV, the linear-cable replacement that recently crossed 8M subscribers (note: at ~$75/mo, subscription fees would add another $7B in revenues; however, the bulk of this figure is paid out to Hollywood studios and sports leagues via the cable networks and broadcasters rather than to YouTube content creators). All in for YouTube advertising, Premium, Music, and linear TV, that’s close to $50B in estimated 2023 revenues. 
 
YouTube’s leading video app generally pays out around 55% of revenues to content creators, which, ceteris paribus, would translate to about $23B of the $42B in ad/sub revenues last year (there could be reasons why this number is lower, but we don’t have perfect transparency from Google on this). That’s an astonishing honeypot that drives creators to, well, record themselves driving, as well as produce a long tail of highly entertaining and informative content. And, the figure is clearly growing as YouTube gains share, particularly in living-room viewing, which demands higher quality/resolution long-form content (that said, even YouTube Shorts are also popular in the living room). Tara Walpert Levy (YouTube’s VP, Americas) noted recently in a Deadline interview that: “The number of creators who are seeing the majority of their watch time come from the television versus other devices is exploding. I think it was up 400% last year. There are some parts of that that we did not expect. For example, Shorts has taken off in the living room.” Walpert Levy also discussed the network effects at work: “Obviously, creators have always been at the heart of YouTube and I think they are now recognized as maybe the no-longer-so-secret weapon in the streaming wars. They’re the studio and the writer and the actor and the director all in one and they’re connecting with people in really new and authentic ways. The second thing is the viewing experience itself. We’re absolutely committed to offering best-in-class products across everything that we do, and to make sure that they’re easy to use and have features this would enhance rather than take away from the viewing experience.”
 
This is, of course, all part of the infinite content trend as video shifts away from increasingly niche Hollywood studios to an array of independent creators (covered in more detail in “Will it Play in Peoria?”). YouTube is 19 years old now, and it’s remarkable to see the role it has played in democratizing content creation and distribution. I’ve held the belief for a couple of years (e.g., see #359) that YouTube is in the best position to bundle all forms of content and drive the highest value for creators and viewers, and this position of strength is clearly growing. And, tagging onto my post last week about the home holodeck, I can’t wait for ambient VR/AR content to enjoy in the future. And, fulfilling my contractual obligation: more and more of this content will be created by AI.

Miscellaneous Stuff
Enzyme to Let You Eat Your Cake Too
A startup is looking to use an enzyme food additive to transform sugar into fiber during digestion to reduce the negative health impacts of sugary foods: “The enzyme Zya is developing comes from a family called inulosucrases, and is naturally made by a strain of bacteria found in the human microbiome that’s capable of converting sugar to fiber in the gut environment. This enzyme acts on sugar before it can be broken down and absorbed by the body. It works by rearranging sugar molecules into inulin fiber, a type of soluble fiber found in plants such as chicory root that fosters the growth of beneficial gut bacteria.” 

Stuff About Demographics, the Economy, and Investing
China’s Export Obfuscation
More evidence is emerging of how US retailers and other corporations are using Mexico to land imports from China, confounding the official stats that show China’s exports declining. I recently covered (#426) a related issue of how the hundreds of billions of dollars in “de minimis” shipments are circumventing US accounting and tariffs; however, these low-dollar direct imports are rivaled by the high volume of all types of overseas commercial goods being re-routed through Mexico. The FT has more stats on containers landing in Mexico from China: 881,000 in 2023, up from 689,000 in 2022 and 600,000 pre-pandemic in 2019. And, Aaron Rubin, founder of one of the largest third-party logistics providers to ecommerce companies in the US, outlines how sellers are using loopholes to land US-destined packages in Mexico and Canada without tariffs. As the FT points out, this changing trade pattern is generally assumed to be a workaround in response to tariffs levied by Trump and sustained by Biden as tensions remain high with China. As I noted a couple weeks ago, the US appears to be as dependent as ever on China, despite the misleading headline numbers showing significant declines in taxed imports.

✌️-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 #427

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: preparing for your home holodeck room; a week of unexpected AI leaps from Sora to Gemini 1.5; sizing the data center market; adding LTM to LLMs; another new Disney droid lands in parks; Billy Joel turns the lights back on and travels through time with AI; and much more below. 

Stuff about Innovation and Technology
Holodeck Rec Room
There is a long history of cultivating dedicated spaces at home for entertainment and media consumption. Even back in Shakespeare’s day, the king had a theater in the palace for live shows (just don’t invite your nephew if you axed his father and married his widow!). The first home cinemas date back to the early 1920s and were a luxury for only the wealthiest of movie buffs. The first modern home theater system was introduced around 40 years ago, resulting in a multi-decade trend of transforming spare rooms or basements into cinematic experiences. All the while, movie theater technology has continuously improved to provide a better experience than home equivalents; and, of course, it also fulfills the social aspect of shared media consumption. Over the last decade, deflation in consumer electronics has made ultra-large screens and good audio attainable for most households; but, after the pandemic bump, the home theater has plateaued in favor of everyone scattered around the house staring at their own phone or tablet. Meanwhile, the availability of content has evolved from expensive VCR tapes to DVDs to streaming to ubiquitous/infinite content from YouTube, social networks, podcasts, video games, etc. And, despite a post-COVID recovery, the movie theater box office has stalled out below its prior peaks. 
 
I think the next decade will be defined by the creation of a new personal media space: the home holodeck. Increasingly, we will want dedicated rooms (perhaps with padded walls!) free of obstacles to experience immersive storytelling, games, fitness, and media. These rooms could also function as social spaces for virtually interacting in a realistic way with people around the world, including attending live events or game play. Multidirectional flooring innovations, like Disney’s HoloTile floor I previously mentioned (#423), and other new innovations may populate our home holodecks. These rooms will benefit from a lack of windows, taller ceilings, and neutral backgrounds. Specially designed, multi-purpose objects will be invented to stand in for virtual objects to increase realism. These morphable shapes and weights could be things you carry, move around, or spar with. The objects might even be interactive, android-like robots that could hand you objects or keep you safe. Imagine self-assembling robotic blocks that create climbable surfaces and changing terrain. Today’s dedicated, decked-out home theaters cost anywhere from $10K to hundreds of thousands of dollars. If you look at the entire spectrum of technology, including the more simple systems of a large screen and media bar, one estimate puts the number of households in the US with some sort of dedicated media room at around 40%. I don't know precisely when or how many households will have a holodeck, but I suspect they will eventually be as prevalent as dedicated media rooms and perhaps even cost a similar amount with some extreme setups that cost many hundreds of thousands of dollars. No doubt there will be an HGTV show about mega home holodeck makeovers in the near future.

The immersive content to populate our home holodecks is already here with devices like the Meta Quest 3, and soon the sky will be the limit. For example, OpenAI’s new Sora video generation engine demonstrates the stunning pace of advancement in AI media creation. You can take a video created by Sora and feed it into a 3D gaming engine, or an app like LumaLabs, and instantly create an immersive virtual world – all emerging from a simple text prompt. It’s a seemingly short step to imbuing this world with rich, complex, AI-driven content and interactions. In another milestone that demonstrates how quickly AI’s understanding of the visual world is progressing, Google unexpectedly released Gemini 1.5 last week, which can process enormous amounts of text, audio, and video.
 
As I wrote a couple of weeks ago in “Your Wish is Granted” and in last week’s Bringing the Virtual to Reality, this is likely a key shift in storytelling that will transform the multi-hundred-billion-dollar entertainment and gaming industry into something very different and much larger than it is today. We no longer appear to be limited by the creative/AI side of the equation; rather, we are limited by the installed base of hardware (e.g., data centers, VR headsets, and the like; also, check out these new AR glasses to get a glimpse into the near future). It took over a decade for smartphones and high-speed wireless to reach 50% penetration. Adoption of the next platform will depend on the pace of innovation, the production of hardware form factors, and the developers who will create all the apps we can’t yet even imagine. At the moment, it all looks to be arriving sooner than anticipated, but timing can be deceiving this early in the innovation curve. 
 
Datalicious
Nvidia CEO Jensen Huang gave an interesting stat on the size of the data center market, and his comments align with my views on the absurdity of trying to scale the chip industry by multiple orders of magnitude: “There’s about a trillion dollars’ worth of installed base of data centers. Over the course of the next four or five years, we’ll have $2 trillion worth of data centers that will be powering software around the world.” By installed base, Huang is referring to all of the infrastructure, networking, buildings, power, etc. that go into a data center, which costs many multiples of the chips that run inside of them. Of course, every dollar spent in the future will buy far more computing power than it would today, given ongoing gains in chip/systems efficiency. Last week, Nvidia surpassed both Amazon and Google to become the third most valuable company by market cap behind Microsoft and Apple. Speaking of historic milestones, a year before I was born, the Warner-Lambert company introduced Bubblicious gum, giving children the ability to blow bigger, more burstable bubbles than ever before. The great thing about Bubblicious is you could blow bigger, more burstable bubbles than ever before – you just never knew when they'd pop!
 
LLMs with LTM
Long-term memory is coming to chatbots, with OpenAI announcing that the feature is currently under trials. As I wrote last July: 
One of the more remarkable leaps with LLMs will come when they have access to long-term memory. [LLMs] have amnesia. As soon as you ask it something, it has no memory of the previous answer or any context through time from other conversations. The entire concept of your sense of self comes from a constantly updated narrative of your moment-to-moment life that you can recall increasingly vaguely over time (combined with being in a body that is taking in sensory data from the world around you). Without this grounding and context, you would have no idea who you are. So, giving LLMs long-term memory could create the ability for them to possess a sense of self and have significantly increased value to users.

Robotic Teams
Disney’s deployment of their new “Indestructible” class of character bots, first introduced in 2023 (see Emotive Bots), is expanding with a Duke Weaselton bot for Shanghai Disneyland’s Zootopia attraction. IEEE reports in detail on the Duke. Due to the instability of bots with more free-form movement, it turns out they work best in collaboration with other robots that can provide support and locomotion. Following the new droids at Galaxy’s Edge, Duke marks the second time in recent history that an emotive robot form factor was rapidly developed and deployed in parks. This raises my hopes that Disney’s HoloTile floors will be available soon!

Miscellaneous Stuff
Artificial Patient Screening
In last week’s Clinical Twins, I mentioned the clinical trial bottleneck that may result as the number of AI-created drug candidates explodes in the near future. Virtual patients can alleviate some of that strain, and a recent study also indicates that AI can greatly automate the manual process of screening to determine the best candidates for a human drug trial. 
 
Piano Man
This video for Billy Joel’s new single “Turn the Lights Back On” features some very well done AI effects that transport Joel on a round-trip journey back through time. The generative AI of Joel aging through his career strikes a poignant tone, especially given the song’s theme. The AI in the music video was generated by Deep Voodoo, a company started four years ago by Matt Stone and Trey Parker of South Park fame. Although Joel has been performing live continuously, his last record was River of Dreams released in 1993.

Stuff About Demographics, the Economy, and Investing
US Labor Force Multipler
The US Congressional Budget Office has increased projected labor force participation from ~5M new workers to ~10M over the next decade. The 5.2M increase – from the forecast from just twelve months ago – is largely due to immigration (PDF). The CBO estimates this big increase in the labor force will add $7T to the US economy through 2034. One estimate puts the 2023 US population growth at the largest on record. This is a welcome offset to the accelerating retirees in the Boomer demographic I wrote about last week, but a rapid growth in the labor force coinciding with rising capabilities of AI and robotics could be setting us up for structurally higher unemployment over the next decade.

✌️-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.