My favorite course at university was Applied Artificial Intelligence. The course followed Stuart Russell’s and Peter Norvig’s Artificial Intelligence: A Modern Approach. I remember reading that book in the library because it was too expensive for a poor student to buy. I also fondly remember sitting many hours in the basement of the university, in front of a Sun workstation, working on an Othello-playing AI agent, and then pitting it against other student’s agents in a tournament.
Fourteen years later, in 2016, I did buy the book. I had just gotten a job in the AI/ML field and thought it would be a good idea to brush up on my knowledge. I also bought a GPU. The GPU turned out to be more useful than the book. Never include the words “a modern approach” in a book title.
In all the years in-between I remained fascinated with artificial intelligence. A machine that could think felt like it would be among the most exciting accomplishments of mankind, perhaps the most exciting. Now that the thinking machines are here I’m having more mixed emotions.
Part of it is that I used to think that artificial intelligence would let us understand what intelligence is; how we think. But The Bitter Lesson thwarted that ambition (just as it favored my GPU over my book). We now have models that can solve a Millennium Prize Problem, but their thinking is just as incomprehensible as that of a human mathematician, if not more so. We’ve managed to transfer the ability of formal thought from the heads of humans to silicon chips, without understanding what it is we’ve transferred or how it works.
Another aspect is that it’s impact started at the unexpected end. I always imagined that AI would first take over the drudgery, the less stimulating cognitive work. But it came first for the mathematicians. I had never imagined that Terence Tao would be among the first to experience the identity crisis that comes from being beaten at your own game by a machine. I thought he would be among the last humans standing.
I’m no Terence Tao, but a large part of my professional identity has revolved around being cognitively capable in mathematics and software engineering. Now those abilities are turning into a commodity. Perhaps AI hasn’t taken my job yet. But it has already taken a big bite out of my identity. That experience makes it easy to sympathize with mathematicians going through the five stages of grief (denial, anger, bargaining, depression, acceptance). The raw talent and herculean effort required to win a Fields Medal is hard to even comprehend. Now a machine can do work on a similar level. That hurts. But at the same time it’s of course an immature reaction. Democratization will always mean that some people who thought they were special lose their privilege. In general that’s a good thing. Even if it hurts.
So what will remain valuable when formal thought is commoditized? Many mathematicians and programmers keep pointing to adjacent skills that models have yet to master as places to retreat to (strongly resembling bargaining). But I doubt those places are safe for long. The models may not be great at explaining proofs or designing large distributed systems today, but they probably will be soon. We now live in a world where the market price of raw intelligence is declining, and it will probably keep doing so. That’s just how it is.
That world is of course also exhilarating. It’s any deep-tech founder’s dream to be able to hire Richard Feynman. Soon that dream may come true for all of us, and the things we will be able to build are staggering. Fields that have been held back by the scarcity of intelligence, or the capital needed to attract it, will open up. Ridiculously ambitious startups will become possible. The challenge then will be to put intelligence to work, to point it at the right problems. One day the thinking machines will surely do that too. Until then, that’s what I want to do.
I’m Björn Smedman, a fractional CTO for deep-tech and health-tech startups. I’m also open to engineering leadership roles in Malmö and Stockholm. If you’re building something ambitious, let’s talk.