SITALWeek #359
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: using language models to program robots and design proteins; robotic pharmacists; robots controlled by VR when they fail; bundling will be king as content becomes ubiquitous; VR can have the same effect as psychedelics; the collision of art and technology might give us a glimpse into the future; the consequences of drought-impacted rivers; Tesla's virtual power plant goes live in California; and, much more below...
Stuff about Innovation and Technology
Powerwalls Ease Grid Strain
During a heat wave in California last week, Tesla’s first test of their virtual power plant system kicked in with around 2600 residential Powerwall owners providing 18MW of power to stabilize the grid. Powerwall owners were asked to opt-in to allow our batteries to kick power back to the grid if needed in return for modest compensation (my offer was "up to $27" per incident). This is only one of many demonstrations that we are at the beginning of a long decentralization of power generation and storage around the world.
AI-First Robots
Google is using large language models, like GPT-3 and PaLM, to help train robots. Wired reports that one-armed, wheeled robot butlers from Google’s spinout, Everyday Robots, can perform such tasks as getting a sponge in response to “I spilled my drink” or a bag of chips in response to “I’m hungry.” Importantly, these are not hard-coded commands with programmed responses like we find in fairly dumb voice assistants like Alexa and Google Assistant. A blog post from Google Research has details on the complexity of teaching a robot to understand meaning behind speech. The ongoing challenge remains: how can we imbue AI and robots with contextual awareness? While they might be able to comprehend a spoken request, there could be myriad missing contexts around the situation, surroundings, and events that lead up to the command. The ultimate goal is for a robot platform or robot operating system to emerge to power a more diverse set of tasks. In the meantime (as noted previously), we have a proliferation of purpose-built machines with little ability to learn and evolve. Approaching robotics from the AI-first perspective, as Google is now doing, may lead to better physical robot design, and we should eventually reach a point where AI robots will be designing themselves using feedback from their ability to respond to tasks. Another robot platform is myBuddy from Elephant Robotics, a cobot based on Raspberry Pi with an emphasis on AI. And, Elon Musk remains committed to producing a multi-purpose, AI-powered humanoid robot, which he sees as the most logical, versatile design for accomplishing human tasks and easily interfacing with society – we will have to wait and see if the AI-imbued bots agree!
AI Protein Design
In other language-model news, we might be able to use AI to learn the “language” of other structures. In one example, researchers report that GPT-2 can design proteins with novel structures/functions based on understanding how amino acid codes translate to function for an existing protein data set. Some interesting potential applications might include designing enzymes for synthesis of pharmaceuticals or detoxification/degradation of harmful chemicals/plastics.
Rxbots
Pharmacists are an interesting example of skilled workers whose job is ripe for automation. As Walgreens raises signing bonuses from $30-50k to $75k to recruit pharmacists from rivals, the company is also planning to get to 50% of prescriptions filled with automated machines by 2025. Mark Cuban’s groundbreaking Cost Plus Drug company is also using centralized, aseptic (i.e., free from contamination) robotic filling systems. Amazon’s PillPack subsidiary is also robot based. A combination of automated filling and data analysis might have avoided the over-prescribing of opioids, which recently cost Walgreens and CVS a $650M judgment in two counties in Ohio. The risk as more and more specialized jobs are inevitably fully automated is the automation paradox, whereby we lose rational human oversight and analog knowhow.
AI/VR Bots
Telexistence is a Japanese robotics company with autonomous robots that can be controlled remotely when they run into real-world problems. They are reportedly working with Nvidia and Microsoft, as well as FamilyMart in Japan, where the company is deploying their TX SCARA robots to stock beverage shelves in 300 of the chain’s 16,000 stores. While working autonomously 98% of the time, they can be controlled by remote workers – theoretically located anywhere in the world to overcome the problem of locally dwindling labor forces in developed countries – using VR headsets for interface/troubleshooting. In what seems to be an increasing trend of VR takeover capabilities, remote workers also interface with Halodi’s security bots, as noted in #354.
Bundling is King
In a world where both content and its distribution are ubiquitous, neither side of the scale can be the proverbial “king”. Content is being overvalued by the creators and producers, and distributors are equally overvaluing their role in mediating the relationship between creators and consumers (see The TikTokification of Consumption). The value in media now lies with the company that can create the highest non-zero-sum bundle of content (likely with both ad-supported and premium options as well as music and perhaps even gaming) and cheaply distribute it to the widest audience globally. Much like the cable companies in the 1980s and 1990s created value by physically connecting homes to a conduit carrying what eventually became hundreds of channels, a company with cloud infrastructure, a very good algorithm, an effective ad platform, and lots of users should be able to create value by bundling today’s popular streaming apps, linear-TV content, and games (much like the excellent, latest Chromecast dongle for TVs, but across multiple devices). The WSJ reports that YouTube is potentially vying for this role as it aims to add streaming apps to the popular platform, which boasts over 2 billion monthly logged-in users and over 1 billion hours per day of video viewing. Although the Hollywood studios are likely very wary of recreating the tension that came about with cable companies, the potential to drive higher ad rates as consumers shift to ad-supported viewing in an ecosystem over-saturated with distribution options might mean the time is ripe to try a bundle. If the studios can maintain relationships with customers and access to data, it would be a far better situation than what they faced during cable’s heyday. Further, with the zealous gutting of broadcast/cable programming and accelerated shift to streaming, which is now 35% of TV time in the US, bundling to reduce viewer churn might become an existentially forced path as overall subscriber growth slows. YouTube, for its part, also holds significant negotiating power as one of the most watched apps across connected-TV platforms. Nielsen reports that, with 7.3% of streaming share, YouTube’s apps are second only to Netflix, with 8% share. YouTube has also quietly built its own powerhouse subscription business, with 50M Premium subscribers as of September 2021, and they recently reported 5M subscribers for YouTube TV (their cable-like linear-TV bundle) on their July 2022 earnings call.
Miscellaneous Stuff
Energetic Coalescence
A VR app, called Isness-D, can reportedly create an experience of transcendence similar to that induced by a moderate dose of psychedelics. The program, which represents 4-5 remotely located participants as dynamic clouds of light, allows users to experience a profound sense of connectedness and ego attenuation. It’s not known yet whether the impact lasts as long or has the same neurochemical changes as psychedelics. Effectively, representing yourself as diffuse energy, lacking definite boundaries between yourself, others, and your environment, appears to alter the brain’s default settings for processing of sensory information, which helps dissolve the discrete sense of self we normally have. One of the most powerful attributes of AR/VR is that it helps us realize that reality is fungible, which in turn helps us understand that the representation of reality our brain generates for us based on sensory inputs is itself a VR experience, having been heavily and subjectively adapted/edited for survival over millions of years of evolution. An AR app I’ve enjoyed in the past that creates a similar sense of connectedness to light and nature is Tónandi, designed by the band Sigur Rós for Magic Leap. The dissolution of our ingrained subject-object duality may prove to be one of the more powerful use cases of VR and AR.
The Terror of Knowing What this World is About
You might have seen clips, popular on social media, from a 1999 BBC interview with David Bowie prognosticating the Internet’s radical impacts on art and society. In the lesser known, full 16-minute interview, Bowie explained how he thought the Internet would transform the relationship between creator and consumer by downplaying the importance of artists as leaders, elevating the importance of the consumer, and causing fragmentation into genres and subgroups. Bowie also said that the Internet would profoundly advance the shift from objective to subjective reality that began in the 1970s. It’s likely survivorship bias – meaning there were probably dozens of other artists that made incorrect predictions about the Internet – but perhaps artists on the leading edge of technology have a special ability to see the present more clearly and divine what the future might bring. A topic we wrote about in our Pace Layers paper was the relatively rapid speed of change in the technology and art layers of the economy. Our paper was inspired by Stewart Brand’s model, which posits some layers of society, like government and religion, move very slowly while others move much faster. Technology has disrupted this model by reaching into the bottom, traditionally glacially moving layers, causing long-standing institutions to crumble quickly. For example, we can no longer take democracy for granted like we might have before social networking took over. And, after all, a social network like TikTok is simply a combination of artists and technology, interacting at an increased pace to impact society. I am not sure where we can look today to find the artist-technologist prognosticators who might give us a glimpse into the future, but they will probably emerge as we shift to augmented and/or virtual reality experiences. Changing the way we perceive fundamental reality is the next step in further subjectifying “truths”.
Stuff about Geopolitics, Economics, and the Finance Industry
NZS Capital’s Acquired Q&A
As part of our sponsorship of the current season of The Acquired podcast (so far covering Walmart and Amazon), we sat down with Ben and David to answer questions about our Complexity Investing framework. You can watch a recording of the chat here on YouTube.
Water, Water, Nowhere
Drought has exposed the economy’s reliance on waterways for both transport and power. The very nature of weather is of course unpredictable, but if year-to-year volatility remains high, we might expect more future instances of rivers rendered unavailable for transport, hydro power, agriculture, drinking water, etc. Manufacturers such as Toyota and battery-maker CATL are temporarily shut down in China due to power shortages caused by extreme drought. The FT also reports that waterways are often used for crop transport. Since one transport ship is equivalent to 100 to 150 trucks, overland options lack the needed capacity to pick up the slack. California is planning for a 10% reduction in water supply by 2040. Hydroelectric energy usage varies widely around the globe, constituting 7% for the US, 17% for China, over 60% for Brazil, and 95% for Norway. One option might be to do more pumped hydro using green energy like solar to move water back up to dammed reservoirs; but, if the rivers are too dry to circularly pump water, it might need to come from the underground water table. The Colorado River, suffering a 23-year drought, will go under restrictions from the Federal government in January that include the state of Arizona reducing 21% of its water use. This should raise some eyebrows as water-intensive semiconductor fabs are being built in the state. TSMC may ultimately use up to 40,000 incremental acre-feet of water a year when fully built out in Arizona, which seems meaningful compared to the 592,000 acre-feet Arizona has to reduce from the Colorado River.
Building for Bots
Pepsi is building a new bottling plant in Denver that is triple the size of its previous site but appears to only need twice as many employees (500 vs. 250). It’s possible that we will increasingly see capacity expansion via new construction of highly automated factories, which can be purpose-built for new technologies to save on labor. This possible revitalization with increased automation in factories comes at a time of burgeoning signs that some capacity is being reshored (see Reshoring Rising). The ongoing labor shortages (for which a likely remedy is currently lacking), will continue to drive automation. Referring back to the opening section of this week’s newsletter, a combination of purpose-built automation and humanoid robots would have a major impact on labor needs over the next couple of decades. It seems inevitable that we get there someday, but there’s a wide range of outcomes concerning what the end-state will look like and how soon it’s realized.
✌️-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.