SITALWeek #418
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 virtuous circle of scientific discovery and AI; Hollywood gets in on the ambient content game; Jaron Lanier on compensating the humans that AI feeds on; reflecting on power laws in the stock market this year; Jon opines on semis. And, much more below...
Stuff about Innovation and Technology
ScienceGPT
What the Internet did for accelerating the velocity and transparency of information, AI will do for scientific discovery. The Argonne National Lab’s supercomputer is working on training a trillion-parameter research-focused LLM called AuroraGPT. The compute system, which has over 80,000 Intel CPUs and GPUs combined, will process scientific data, research, and papers with the intent of creating an assistant to speed up basic research. Microsoft is working with the lab to scale up the training to all 10,000 nodes, which could take several months to train to completion. Meanwhile, DeepMind is releasing the models of several-hundred-thousand novel crystalline structures that could revolutionize fields like solar panels, superconductors, batteries, semiconductors, and more. The trove would have taken humans 800 years to discover at our pre-AI pace. The real wild part comes when you start to feed this type of data into something like AuroraGPT to identify potentially viable inventions and breakthroughs for humans to further explore. Researchers at Berkeley have already taken some of the suggestions from DeepMind and used an LLM to suggest synthesis recipes for 58 target materials. Using an automated lab, 41 of the targets were created. The abstract of the paper concludes: “The high success rate demonstrates the effectiveness of artificial-intelligence-driven platforms for autonomous materials discovery and motivates further integration of computations, historical knowledge and robotics.” While I often talk about the long time lapse between a digital breakthrough and its impact on the physical world (see When Positive and Negative Feedback Loops Collide), it’s probably time to start thinking about a faster pace of disruption and innovation affecting a broad swath of traditionally slow-grinding economic gears. The Internet and smartphones created digital advertising, digital media, ecommerce, cloud software and more, but these will likely all pale in comparison to the accelerated scientific revolution coming to the physical world of materials and biology.
Lanier’s Lens
This brief interview with VR/AI-pioneer Jaron Lanier has a few interesting bits on generative AI. While I am generally in favor of editing, unfortunately the video appears to have been painfully edited to cut away right when Lanier was about to get to the interesting stuff. I think there are some people who should have an uncompromised license to expound, and Lanier is one of them. Around ten minutes in, Lanier talks about the potential to compensate people for contributing to AI models via attribution. The slightly awkward thing here is that Lanier works for Microsoft, which has the potential to use its new copilots to train on people who then will lose their jobs to those same AI copilots. Lanier’s comments could be construed to argue that Microsoft should be paying Windows users to train their copilots, and then giving them royalties in perpetuity. What are the odds of that happening!? And, despite working closely with Sam Altman and OpenAI, Lanier is quick to challenge Altman’s Worldcoin business, which Lanier characterizes as: “Sam wants to do this thing of a universal eye scan based cryptocurrency coin to reward people once AI does all the jobs; obviously that's not within my frame of recommendation. I don't think that's a good idea. I think that some criminal organization will take that over no matter how robust he tries to make it. Like look at crypto, crypto is mathematically perfect and then at the edge it's all criminals and fraud and incompetence.”
Doing More with Less
Last week, I mentioned that Slack had, at one point, tried to subsume the user interface for enterprise software apps, and I voiced my view that chatbots/copilots will all eventually just talk to each other, cutting out the humans in the middle. I came across an interview with the Slack CTO and co-founder that discussed Slack’s plan to integrate more AI models: “So I think almost every kind of information worker role is going to be augmented by AI in some way over the next couple of years. And we're really just starting to scratch the surface. Of course, because these tools are being built by software engineers, software engineering is the first area that's augmented by it. But I think we're going to continue to see it across a wide variety of disciplines of taking some portion of that work, and being able to automate it, make it faster or accurate, and leave time for the higher leverage work.” That last bit about freeing up time sure sounds like a recipe for doing more with fewer employees, similar to what we’ve seen in software engineering’s early adoption of AI copilots. Cynicism about job losses is likely to lose out to optimism long term as AI creates a larger economy with more opportunities, but it’s getting harder to be optimistic about the prospects for many rote, white collar, computer-based jobs over the next few years. (I hesitate to link directly to this interview because the website appears heavily spammy on ads and requires readers to opt into personalization just to see the article, but if you want to track it down it’s with Cal Henderson on a site called Moneycontrol.)
Miscellaneous Stuff
Ambient Hollywood
In #404, I joked that Disney’s Park visitors uploading walking tours to YouTube were creating competition for Disney’s expensive movies and streaming shows. Well, not to be outdone, Hollywood has been getting in on the ambient video business over the past year. On Paramount+, you can chill to three hours of the Star Trek Enterprise NCC-1701-D warp core or watch a crackling fire by a lake in front of Paramount Mountain. Max, meanwhile, is carrying a chill, background music/animation show called [ambient swim]. And, on Disney+, you can now find many Scenescapes Ambiance channels with sounds and backdrops from the Haunted Mansion, Star Wars, Dory’s Reef, and Arendelle Castle. Perhaps these low-budget programs are all preparing us for our inevitable future of 24/7 AI-generated ambient content.
Stuff About Demographics, the Economy, and Investing
M7 Power Law
One of the key tenets of Complexity Investing is that we cannot reliably predict the future; therefore, we instead focus on fundamental characteristics of businesses like adaptability and non-zero sumness to assess investment potential. Likewise, there is very little, if any, useful information to be garnered from current market asset prices, as those numbers generally reflect an array of opinions from countless people, and increasingly AI, all based on chaotic inputs to arrive at a random price that reflects supply and demand in the moment. The best we can do as investors is to rely on as few predictions as possible and a handful of truths from complex adaptive systems to suss out potentially interesting assets. That leaves the final piece of the investing puzzle: when to join those assets on their journey through time and when to leave them, i.e., when to buy and sell. This last bit complicates the picture greatly, of course, and it opens the door for a wide variety of cognitive biases and compounding mistakes that build on each other over time. Unexpected market events occur with marked regularity thanks to the emergent behaviors of complex adaptive systems. But, occasionally, a trend stands out against the normal noise of ebbs and flows. Right now, for example, there are a very small number of very large companies that are the primary drivers of the global markets. This year, the MSCI ACWI, a broad measure of the global markets, is up 16.74% on a total return basis through November 29th (all numbers below are through the same date). In particular, the growth-focused version of the MSCI ACWI is up 27.68%, while the value version is only up 6.19%. That divergence is not atypical, but let’s drill down further. The US has fared better than the global index so far this year, up 20.29% (using the S&P 500 as a proxy). We can use the Russell 1000 Growth Index (R1G) as a proxy for US growth stocks (the index includes approximately 450 larger companies that are assumed to have higher valuations and higher growth levels), and we see that it’s up 36.58%. That is six times the value index and more than twice the overall return of the global markets. To understand why, let’s look at the so-called “Magnificent Seven”: Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia, and Tesla. The M7 are up 101.42% so far this year! These companies, as of November 29th, had a total market value of just under $12T, representing around 40% of the R1G (according to Bloomberg, the market capitalization of the R1G was $29T as of November 29th). To put even more context on the number $12T, the Russell 2000, which represents the next 2,000 largest US companies just behind the top 1,000, has a total value of just $2.72T. Those 2,000 companies are worth less combined than the current value of either Apple or Microsoft.
A little bit of contribution math tells us that the M7 accounted for roughly two-thirds of the R1G performance this year. Similar math emerges for the S&P500. Various reports indicate that this is the narrowest rally ever for a market that has risen more than 15% in a year and this is the most seven stocks have accounted for the market capitalization in history. We expect power laws to be a natural occurrence in complex adaptive systems, but this one is attention grabbing. To have all of the very largest companies all lead with some of the best performance, in some cases for the same reasons, is certainly not something that happens every year. I fear I will disappoint you by not commenting on what it all means or where it goes from here. SITALWeek has covered the salient trends of the analog-to-digital transformation of the economy (including regulatory capture, network effects, power laws, etc.) extensively, and, clearly, much of this outperformance is understandable through that lens. AI, in particular, has the potential to create larger power-law winners, while at the same time increasing the fatness of the tail to the downside risk outcome as well. So, perhaps the extreme market power law of 2023 is explainable from this perspective, or perhaps we should just go with the obvious joke: the AIs are all influencing the market algorithms to drive up the share prices of their creators! One thing we know about technology cycles is that they can be in one of three phases: overspending bubble, bursting bubble, or post-bubble-burst normalization. At the moment, we are somewhere on the tenuous ground between the first and second phase, but these things are unpredictable and defy the odds for long periods of time.
Semi Summary
Jon recently gave a talk to the CFA Society of Switzerland on the semiconductor value chain and related geopolitics. You can see the presentation here.
✌️-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.