SITALWeek #369
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: Halloween is a big holiday at our house, resulting in an abridged topic list this week. I explore a couple of ideas around "generative AI" – the trending term for the many transformer AI models that are rapidly becoming a hot area of focus for many industries, most notably design and art; a look at the folks trying to blend cultured meat with plant material to recreate a tastier and more economical animal-protein substitute; and, lastly, an interesting look at how information consumption varies at companies, which follows a pattern we expect to see in living organisms. If you missed last week's much-discussed exploration of the perils of algorithm use by corporations, and how it likely caused the crazy spiral in apartment rents, check it out here. Have a great Hall🎃ween!🤘
Stuff about Innovation and Technology
AI Copycats
The field of generative AI, the term that encompasses text-to-image, audio, and video creation/manipulation algorithms driven by short user prompts, is advancing rapidly. As new platforms proliferate, the concern over copyrights for AI art is rising, and the Recording Industry Association of America is worried that generative engines will be trained on their recording artists’ music without their consent/compensation. There is a common process in the music industry for paying royalties to sample copyrighted music, which would be very hard to follow if AI engines can’t trace their sources. CNN also spoke to several painters who were upset to see AI models easily recreate their painting style after being trained on their images. Stability AI, the startup behind Stable Diffusion that recently raised money at a $1B valuation, told CNN it’s working on ways to potentially compensate artists. One of the big issues today is artists aren’t even being asked permission before their works are included in training. I’ve covered the topic of AI art recently, and it seems clear these engines will need to be able to list the works that inspired the final product. It’s a rather tricky situation, but it could be solved (at least partially) by artists being able to copyright a “style” rather than a specific artwork itself. There may be some precedent in other industries; for example, car makers have a history of suing each other over things like similarly styled front grill designs. But, where do you draw the line for commercial art? If you see a painting or hear a song that appears characteristic of a specific artist, maybe that “style” will/should be copyrightable by the original artist (or whoever owns the rights).
Artificial Homework
Back in #355, I covered the rise of text-generative AI, such as Sudowrite, for assisting authors with book writing. Last week, The Information reported on the increase in students using such tools to write and edit papers. From high school to college level, many kids are getting good grades while the AI does the work. One student “felt confident in GPT-3’s ability to complete college-level work, because it had helped him graduate from high school. For one biology homework assignment in 12th grade, he was asked to write a paper describing all the functions of a cell. ‘I just really didn’t want to do it,’ he said. So he gave it to an artificial intelligence writing tool to produce. He looked at the text the AI generated and felt it was good enough to submit—so he did. His teacher gave him 100%.” Jasper is an example of text-generating AI, while others, such as QuillBot, edit what you have already written. These tools are yet more examples of the swiftly evolving nature of creative output. And, the next logical step might be to embed some of these AI assistants into word processing programs, rapidly amplifying accessibility and training of the AI models. Will we soon have a plethora of Stephen King wannabes churning out 1000-page horror crime thrillers with little more than a handful of specified plot points? As noted above, authors may want to investigate copywriting their “style” as well.
Miscellaneous Stuff
“Hybrid” Burger
In a section of #316 titled Cultured Meat Pie in the Sky, I noted the massive amount of bioreactor capacity and other hurdles that would need to be overcome to make even the tiniest of dents in the meat industry with lab-grown animal protein. And, since the current cadre of plant-based burgers doesn’t appear likely to win over a world of meat eaters, some companies are taking a different approach to meat replacements. A startup called SCiFi Foods is attempting to blend cultured meat cells with plant-based meat alternatives to create a cost effective, yet tasty, alternative to farm-raised meat, according to Vox. There remains wide skepticism, but optimistic estimates suggest that this hybrid approach could be competitive with traditional meat by the end of the decade. Another startup, New Age Eats, includes cultured animal fat cells (in addition to muscle cells) in their hybrid products, which is reportedly key to taste and mouthfeel in replicating real meat. If all goes well, maybe these new meatier plant burgers will be served up by the RoboBurger robot. The company, which recently raised $10M in new funding, has an autonomous machine that can cook and assemble a burger in four minutes.
Stuff about Geopolitics, Economics, and the Finance Industry
Information Footprint
According to a new paper (pre-print PDF), information consumption at companies follows a 3/4 power law, similar to scaling trends found in complex systems and biological organisms throughout the world, from circulatory systems to cities. Here is an excerpt from #315 describing the basic concept:
We’re big fans of Geoffrey West’s work and have often recommended his book Scale to people who ask us about complex systems and the Santa Fe Institute. In Scale, West describes how physiological characteristics of mammals follow quarter power scaling laws. For example, an elephant weighs 10,000x more than a squirrel and grows 10x slower, has a 10x longer gestation period, and lives 10x longer (10 being 10,000^¼). Because growth is dependent upon distributing nutrients to cells throughout the body, one might think that volumetric (a.k.a. weight in this case) differences would determine growth rates, giving third power scaling (since volume is three dimensional), which would correspond to 22x slower growth/gestation. Why is the elephant able to eke out a faster growth rate, and thus better reproductive fitness? The fascinating reason for this quarter power scaling is likely due to blood vessel and respiratory networks that have evolved through natural selection to maximize metabolic rates in the three-dimensional world we find ourselves in. The best way to maximize space in a 3D world is to use fractal, or self-similar, patterns. Fractals are so efficient that they essentially give a network, like blood vessels, an extra dimension, which is why the number four repeatedly shows up as a scaling factor – it’s three dimensions...plus one. In other words, a linear fractal is two dimensional, a surface area fractal is three dimensional, and a volumetric fractal is four dimensional.
This fractal upscaling might also explain the researchers' findings – that larger firms superlinearly consume more information (i.e., news articles) per employee than smaller firms – implying that “The tools of the knowledge economy enhance firm productivity” (p. 7). If we extend the fractal analogy, I might surmise that firms are maximizing the number/depth of topics they explore relative to their constrained size. Of course, it’s possible the scaling effect is a coincidence, but, given large organizations can be viewed as living organisms, this interpretation seems plausible. Curiously, reading amount was more closely tied to a firm’s revenues rather than number of employees, again implying some benefit to economic size. And, the common path of many firms “suggests that large public firms are increasing[ly] confined to a deterministic trajectory in terms of information diversity” (p.4). Firms that start out with very specialized information interests are slower to diversify topics. An interesting hypothesis that comes to mind is whether the type of information diet (diverse, connected, etc.) a firm consumes directly impacts the adaptability/success of their fitness function (which defines their operation within their sector). Specifically, would a more information-diverse organization see disruption coming before a competitor? The data in the article were limited to a small window, but it would be interesting to trace the metrics of information consumption over a longer span (5+ years) to find correlations with expansions, revenue growth, vertical integration, self-disruption, or entry into new sectors. If anyone has that type of company-wide (anonymized) article access data they would be willing to share, please contact me.
✌️-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.