Is NVIDIA the New General Electric?

And Peter Ward On The Extinction Of Life On Earth

My Most Recent Podcast

My most recent guest on Brave New World is paleontologist Peter Ward. Peter is a professor at the University of Washington and the author of numerous books, including Rare Earth and A New History of Life. Given the vastness of geological time, which spans billions of years, I’m blown away by how precisely Peter is able to describe the ten extinctions in Earth’s history, and explain the resurgence of life following each one of them. By precisely, I mean to within a few thousand years. And it is sobering to read that life on earth is headed for ultimate extinction, and to think about ways we might gain a few years in the meantime.

I asked Peter what distinguishes life from non-life and how life might have gotten started. You’ve got to listen to his full answer, but I’ll highlight one definition he attributes to NASA: life is a substance that creates “negative entropy,”  and is capable of evolving and reproducing. Entropy means disorder, or randomness, so negative entropy means its opposite, order. I’m a chemical engineer by training, and I still remember the laws of thermodynamics, one of which says that the entropy of the universe is increasing. Life somehow defies disorder by creating order using energy.

It made me think about how Artificial Intelligence is also all about finding order from disorder. Machine learning algorithms extract order from their training data. Machines could also reproduce. So, could we think of AI as the basis for artificial life?  Could we, that is “wet life,” just be the engineers who produce the artificial life that becomes the more enduring life form after we eventually disappear from Earth? A few years ago, this would have sounded like science fiction. Now it seems like a matter of time.

It’s an eye-opener talking to far-frontier kinds of scientists like Peter. So, check out the conversation:

https://bravenewpodcast.com/episodes/2024/01/25/episode-76-peter-ward-on-life-on-earth/

Is Artificial Intelligence a General Purpose Technology?

I just returned from a two week trip to the West Coast, where I was teaching a “Tech Innovation” class in our Tech MBA program at NYU Stern. Between classes, we visited about a dozen tech companies: behemoths like Microsoft, Salesforce, Amazon, and NVIDIA, large companies like AirBnB, LinkedIn, and T-Mobile, medium-sized companies like Palantir, and innovative startups such as Saildrone and Upside foods. Saildrone makes solar powered autonomous sailing vessels, and Upside Foods makes cellular meat. I’ve given up meat, but I just had to try their chicken tacos. They tasted very good, but I was hoping to sample a medium-rare steak. Maybe that will be ready by next year! (I had a great conversation with Paul Shapiro of The Better Meat Co. on the nascent state of the cellular meat industry over a year ago, so it was interesting to see the progress we’ve made since then. It’s impressive.)

As everyone can now plainly see, AI is everywhere. A question I posed to the class is whether AI has gone from being an application to a “general purpose technology.” I make a case for it in my article “Intelligence,” where I argue that the emergence of pre-trained models provides “General Intelligence” which is configurable for any purpose. My view is that the latest paradigm shift makes AI a general purpose technology and a commodity like electricity, which pervades all applications and whose quality keeps improving as increasing amounts of data and computing power become available.

Who wins when general purpose technologies emerge? Does electricity provide any clues? Who will be the winners if AI blossoms as a general purpose technology, transforming the economics of virtually every industry, including entertainment, finance, transportation, and health?

The Winners

I teach a class on Systematic Investing based on my experience using AI for trading in financial markets. I believe that on a daily basis, data-driven algorithms are better than humans at prediction. But I have my favorite humans, such as Warren Buffett and my colleague Aswath Damodaran, who bring something special to prediction: discipline and a process. I’m also impressed by Tetlock’s “Superforecasters,” who follow a rigorous process that enables them to focus on the relevant information and ignore the chatter. I’ve had great conversations on prediction with both Aswath and Phil Tetlock, which I have listened to several times because there is so much in them.

General Electric (GE) was the dominant player for almost a century following the commercialization of electricity in the late 1800s. It went into all kinds of adjacent industries that involved the production and use of power, and dominated the business landscape for the next century until it unraveled over fifteen years ago. NVIDIA could well be the next GE.

I became an NVIDIA fan after I attended NVIDIA CEO and founder Jensen Huang’s talk at NYU’s “The Future of AI” event in January 2016.  Like most people, I’m not great at stock picking, but NVIDIA seemed very well positioned for AI, a space I have been in since my doctoral student days. I mentioned it to my Systematic Investing students as a low risk play with massive upside potential. “Deep Learning” was emerging as the dominant AI paradigm, and NVIDIA’s programmable GPUs had demonstrated tremendous potential for parallel computing which the new paradigm required. Since then, I have argued that the emergence of pre-trained models such as Large Language Models (LLMs) that can be configured for all kinds of future applications have transformed AI into a general purpose technology.

I discussed NVIDIA as a case study in my Tech MBA class prior to visiting the company last week in Santa Clara (a big thanks to Stern alum Rama Akkiraju for hosting us). Having enjoyed the ride up in NVIDIA stock, I’ve been asking myself when I should sell. After all, no company dominates forever. Indeed, in June 2023, Aswath did a brilliant analysis on NVIDIA, ending with “I would be lying if I said that selling one of my biggest winners is easy, especially since there is a plausible pathway, albeit a low-probability one, that the company will be able to deliver solid returns, at current prices.” The post has over 26 million views, so when Aswath talks, people listen.

I chatted with Aswath a few days later, observing that his musing had sent NVIDIA stock down almost 10%, but I wasn’t quite ready to unload. In my podcast conversation with Aswath in March 2022, where NVIDIA also came up, he mentioned that one of the downsides of his kind of value investing is that you tend to sell your biggest winners early because expectations get untethered from the fundamentals.

On the bus back from the NVIDIA visit last week, I asked my students whether they would be sellers of NVIDIA stock at the moment. Is it positioned to provide the AI horsepower that will drive innovation in all kinds of industries, like GE did with electricity?

The consensus was that as long as Jensen Huang is in charge, don’t sell. I’m going to follow their advice for now.

Electricity created numerous winners. AI is likely to be no different. Exactly a year ago, after my west coast Tech MBA class, I wrote about how AI will kill car brands, and I expected NVIDIA to be front and center of that impending disruption. My logic was simple: cars will consist of the body and the brain, and the real value will be in the brain, which NVIDIA will provide, while traditional automakers fight over the crumbs building commoditized bodies. I still hold that view. But I think that automobiles will be just one of many industries that will be disrupted by AI, so if anything, I may have underestimated the future impact of AI and its potential market size. It’s going to be a fascinating year ahead.

Who the other winners in AI will be in addition to NVIDIA is a great question for Damodaran and for Tetlock’s superforecasters, one that I hope to discuss with them in the months ahead. Stay tuned. We sure live in interesting times.