Adapters, Prediction Markets, Groq

Disrupters, Adapters, Victims

Generative AI has replaced interest rates as the top factor moving markets. It’s not a coincidence that OpenAI, Google, Meta, NVidia, Apple, Amazon and others originated in the US. US equity markets have been the dominant beneficiary of the AI wave to date, and that will likely continue. The wealth gap will widen as small elite tech teams achieve new economic heights previously thought unattainable. AI and machine learning are fundamentally a positive sum game for society.

It’s not a good time to be agnostic to the generative AI and machine learning trend. We’ve written in the past about disrupters, adapters and victims. The disruptors are reasonably obvious; they move fast and break things. They are often smaller by nature because they are new. The adapters are incredible stories of companies like Deere, known for its tractors, which evolved from its founding in 1837 before the discovery of gasoline to today with its own constellation of satellites to help with robotic farming machinery and remote sensors. The victims are the ones who fail to adapt to the new regimes as they come, they often don’t even attempt. In a few years, generative AI and machine learning functions will take over just about every application, and the companies that are not actively experimenting and implementing their vision of the generative future will be left behind. Investors must be tuned into the evolution of the disrupters, adapters and victims in their portfolios, whether it’s exposure through funds or the individual companies that make up their portfolios.

The biggest risk is not taking big risks. The experimentation feedback loop is now intensely tight as the timeline of an idea leading to the first one million customers is shrinking. I’m curious to see how the biotech sector disrupts and adapts going forward. Transformer models used in ChatGPT successfully deduced grammar rules when trained on a complete collection of the written language. Once large models are trained on raw DNA ATCG data and other linked biological markets, there is potential to discover similar “grammar” rules for life sciences. For now, it seems that a lot of biotech operates in a massive labyrinth of opportunity and dead ends. It will be an exciting time when those doors are unlocked, as there really is no bigger market in the world than health. Health-related AI breakthroughs exemplify the broad positive-sum outcomes I mentioned above.

Back to interest rates, the Fed has to contend with another bifurcated economy, similar to the “K” shaped recovery post-Covid. On one side, we have a white-hot exponential US tech sector unphased by 5.5% interest rates. On the other side, we have serious global slowdowns, notably in China and Germany, inflation hitting many domestic small businesses and other countries, extensive commercial real estate issues and broad confusion as to whether the US consumer is and will be financially ok. Aside from tech, most of the stock market has not performed well over the past 12-18 months. Higher stock markets add to consumer and economic confidence. Larry Summers has now suggested that another hike might be on the table. That’s highly contrarian and unlikely, but it’s good food for thought.

Prediction markets

Online unregulated prediction markets like PredictIt, Manifold and Polymarket are interesting for their data on controversial topics. Seeing how people allocate money based on their convictions is more interesting than their opinions alone. There is also a relatively new paid professional prediction company called Infer, which is owned by the policy non-profit RAND.

All markets come with their strengths and their quirks. My favorite quirk from recent memory was how wrong the market was on Polymarket when it bet that Sam Altman would be the Time Magazine 2023 Person of the Year. It ended up being Taylor Swift, to no surprise from the general public. Prediction markets are often quoted in media and financial presentations. I’ve seen Blackrock presentations that quote PredictIt stats. Consider the biases of the participants in these markets when scanning probabilities.

Groq

Groq’s LPU (language processing unit) has gained attention as a true competitor to NVidia’s H100 GPU (graphics processing unit) used in AI applications. There are a few things to note here. Groq’s LPUs were fabricated in the US, whereas all other high-end chips, including NVidia’s, are made in Taiwan. Two, Groq was founded by one of the original transformer paper authors published at Google in 2017 which opened the door for ChatGPT and other models. I mentioned last week that the talent pool in this industry is small and in high demand, which could lead to engineers leaving to start their own companies. Third, NVidia’s B100 chip is expected to be announced next month which could be another leap forward for the NVidia product suite. These are highly specific details but the evolution of this space will have a big impact.

Meta glasses

Meta has another augmented reality glasses unit called Aria. I briefly tested a friend’s Vision Pro headset last week; it’s amazing, as expected. Interestingly, Meta has both the Quest 3 (their less expensive immersive competitor to the Vision Pro) and this test product, Aria. I would use a lightweight product format like these glasses over the ski goggles format. I suspect many others would as well. I hope Aria is released soon so that we can see what this format is capable of. Eye tracking is an interesting substitute for using a mouse that I hadn’t considered, and I wonder if that alone would make it worth using the product. Once you try the Vision Pro, you’ll know what I mean about a mouse substitute. The closest current competitor I’ve found in this format is the XReal Air, which functions as a large screen.