SITALWeek #402
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: smashing printers; companies and workers under threat by AI should embrace and experiment; the Gordian knot of AI copyrights; GLP-1s' desire-zapping capabilities; pirate otters; little government division that enforces the chip war; and, much more below.
Stuff about Innovation and Technology
Preempting Printer Pounding
We’re all familiar with the infamous printer destruction scene in Mike Judge’s 1999 cubicle-culture movie Office Space. Even the most equanimous Buddhist can be incited to rage by a persistent printer error. The Washington Post reports on the rising popularity of so-called smash rooms around the US where you can clobber a variety of items, the most popular of which is...printers. Unfortunately, local party-pooper regulations are popping up, preventing printer pummeling due to dangerous chemicals, batteries, etc. in the poxy appliances.
Shutterstock’s Attempted Goldilocks
For many companies, AI will be both a threat and an opportunity, and determining the right combination of offense and defense will require experimentation and adaptability. Shutterstock is an interesting example of a company facing a series of AI Catch-22s that needs to find an AI porridge that is just right. The leading stock photo company embraced AI early, licensing its library of images to OpenAI for training back in 2021. They also allow their contributors to upload AI images for their marketplace, and a recent deal expands the relationship with OpenAI for training on images, videos, and music in Shutterstock’s library. In return, Shutterstock’s contributors can use an embedded version of DALL-E to manipulate stock images (or create their own work from scratch) and sell the company their AI creations. However, ultimately, platforms like Midjourney (or the array of new features in apps like Photoshop) can turn anyone into a stock photo creator. Feeding the beast of LLMs by licensing your stock photo and video library is a bit like biting the hand that feeds you, because your contributors might ultimately be replaced by AI, and it’s also feeding the mouth that bites you, because AI might end up eliminating your entire business. At this early stage of generative AI, I suspect Shutterstock’s approach to embrace and experiment is the right choice, but it’s clearly not without risk. Long term, contributors and artists need to stay a step ahead of AI, finding new ways to add value, and marketplaces like Shutterstock should disrupt themselves before they are involuntarily obsoleted. The online publication Rest of World recently did an in-depth look at the impact of AI on gig jobs around the globe, and the transformative power of using generative AI for custom images was one of the most striking examples. As one profiled worker said: “You shouldn’t be afraid of AI, you should be afraid of people who use AI to make their work even better”. Given the complex outcomes that can result once AI gets involved, at the very least, the range of outcomes is widening for industries fueled by monetized images/videos. One lens we always use at NZS Capital is looking for ways to maximize non-zero sumness, or win-win for all constituents. In this case, OpenAI clearly benefits, Shutterstock benefits from getting early access to OpenAI’s image generation tools and from licensing revenue, and contributors can benefit as Shutterstock negotiates on their behalf for royalties if their creations are used by AI. However, the artists likely face the biggest Catch-22 here and may need to evolve how they add value in the creative process.
Litigatory Distraction
Other uses of creative works by AI remain uncertain. Comedian and author Sarah Silverman recently sued OpenAI and Meta for their purported use of her books in their model training without permission or credit. Silverman is joined in the suit by authors Richard Kadrey (of the Sandman Slim series) and Christopher Golden. The complainants believe these two companies obtained bootleg digital copies of their books from torrent sites that were used to train the LLMs, which, when prompted, will offer summaries of the books (however, given that online book reviews are common, not to mention welcome by authors, it seems possible that these AI models may have simply crafted summaries from reviews). I am of two minds on this issue. On the one hand, I think creators should be compensated for copyright material based on the laws for each medium, whether it be text, music, patents, etc., especially if their work is used in a commercial manner. On the other hand, I think LLMs are human-like, so perhaps behaving like a legally abiding person should be permissible. For example, it’s a safe bet that Silverman has sat in comedy clubs and taken bits and pieces of jokes and/or joke structure from other comedians, either consciously or subconsciously, and then processed that knowledge in complex ways with new inputs from her life to create new jokes, which she can ultimately perform in lucrative streaming stand-up specials. From this narrow view, Silverman doesn’t seem too different from an LLM, and she is not required to specify and provide compensation for everything that might have led to a seemingly novel punchline. Indeed, I can make this argument for all human creations and endeavors, which would imply that there is nothing truly original since it’s all built on prior works. Afterall, what is every novel but a variation of Shakespeare, which was a variation of Greek myths, etc. Obviously, this extreme lens is not practically useful, and a line needs to be drawn somewhere to allot credit where it’s due. However, it seems unfeasible (e.g., from a time and energy perspective) to require LLMs to reconstruct their logic in minute detail (and what would we do with the reams of output, feed it into another LLM?). I couldn’t retrace the mental gymnastics of how every little thing I read or heard contributed in some way to SITALWeek even if I were to take copious notes, that’s just not how the human brain – nor perhaps LLMs – are designed to work. I think today’s sophisticated copyright and fair-use laws should be updated to reflect the new LLM army of human-like brains consuming content and creating new outputs. Currently, output from AI is not copyrightable, which doesn’t seem right given the high degree of artistry/creativity exhibited by the human prompter (and, one might argue, the AI itself) in creating some of these works. Ideally, artists should use these tools to write even funnier jokes, create even more amazing art, invent new products, or make amazing discoveries rather than sue over book summaries that can turn up in any web search today. The more depressing angle for us humans is that we are already awash in infinite content, and it’s about to explode to unimaginable levels. It seems like a less than ideal use of time for creators of that content to be arguing over this stuff, rather than focusing on how humans can continue to stay a step ahead of the capabilities of AI.
Miscellaneous Stuff
Desireless Diet
Back in #397, I noted the growing evidence that the new class of weight loss drugs, GLP-1s, might also curtail other addictive behaviors besides eating. A new story on STAT explains further the mechanism by which GLP-1 agonists alter the pathways in the brain that drive desire for all sorts of things that may or may not be great for us: “The exact mechanism by which these drugs cause weight loss is still murky, but the bulk of the data points toward this explanation: The gut and the brain both make GLP-1. In the gut, it regulates metabolism — turning up insulin while dialing down glucagon, a hormone that stimulates the liver to break down carbohydrate stores into glucose. Independently, the brain uses the GLP-1 for other things, like regulating energy homeostasis, activating stress responses, and diminishing reward drives. GLP-1 from the gut doesn’t usually get into the brain. But GLP-1 agonists do.” As I noted in #399, there might be a link between these agents and a general lack of desire, which could help explain why people find it hard to stay on their prescribed drug regimen, i.e., a loss of interest in pleasurable things is hard to, well, stomach. As LLMs become more human, it’s hard not to envision that these desire-zapping drugs are making humans more like LLMs.
Furry Board Thief
Amongst a rise in recent out-of-the-ordinary aquatic mammal behavior, in Monterey Bay, authorities are on the hunt for an aggro sea otter that continues to steal boards from surfers in an apparent bid to hang ten.
Stuff About Demographics, the Economy, and Investing
Big Little BIS
If you are wondering which part of the US government is in charge of the chip war’s frontline, it’s the Bureau of Industry and Security, a small division of the (likewise diminutive) Department of Commerce. BIS processed 100K export licenses a year at the height of the Cold War, according to the NYT. Reduced regulations subsequently allowed the number to fall to 10K, but global tensions have again ratcheted licenses back up to 40K. Despite a dearth of current controls and the difficulty in tracking chip sales (given the lack of proper KYC or software solution), the BIS has only three enforcement agents on the ground in China. As I’ve noted in the past, AI-driven advancements in software efficiency may negate the need for leading-edge chips, which would make the chip war moot. Regardless, having a small division of a small department in charge of one of the supposedly most important issues of the moment doesn't seem to add up.
✌️-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.