Been meaning to write about this for almost two weeks now and kept getting distracted (the usual: a client site went down, then my cat knocked a full glass of water onto my keyboard, RIP a couple keys that still stick a little).
So: Geoffrey Hinton won the Nobel Prize in Physics this month. Split it with John Hopfield, "for foundational discoveries and inventions that enable machine learning with artificial neural networks." The Royal Swedish Academy announced it back on October 8th, and I've been chewing on it since.
Here's the thing that gets me. Hinton is the guy who quit Google in May last year specifically so he could talk openly about how worried he is about the stuff he helped build. He's done the interview rounds, the "I regret my life's work" quotes, the whole bit. And now the same body that hands out prizes for, you know, actual physics discoveries has decided his backpropagation-era work belongs next to superconductors and gravitational waves.
I'm not saying he doesn't deserve it. Backprop and Boltzmann machines genuinely reshaped an entire field, and Hopfield networks are the kind of foundational thing you don't appreciate until you try to explain gradient descent to someone and realize half of it traces back to work from the 80s. I took a machine learning elective in undergrad that spent an entire lecture on Hopfield networks, and I remember thinking at the time it felt almost quaint, this weird associative memory model, not the thing that would eventually become the plumbing under half the internet.
But the prize lands in this strange spot. It's not really a physics prize in the way anyone in 1990 would have recognized. It's more of an achievement award for a field that didn't fully exist as "AI" when the original work happened, retrofitted onto a physics category because there wasn't anywhere else obvious to put it. I get why they did it. I'm just not sure "physics" is doing any real work in that sentence anymore.
And then there's Hinton himself standing at the center of it, somehow both the guy being honored for starting the fire and the guy standing outside yelling that the fire is getting out of hand. That's not really a criticism of him, to be clear. He's been pretty consistent about it, going back to that whole thing last year. It's just a strange kind of prize to receive: applause for something you spend most of your public interviews expressing genuine unease about. I don't know if there's a precedent for that. Maybe Oppenheimer, if you squint, though that's a comparison that gets thrown around so much at this point it's basically worn smooth from overuse.
What I keep coming back to is how normal all of this has started to feel. Five years ago a Nobel for neural network research would have seemed like a stretch, a nod to a trendy subfield that might not pan out. Now it's just obviously correct, almost overdue, and nobody I know even blinked at the announcement. That's the actual story here, more than Hinton specifically. The stuff that used to be the fringe, slightly embarrassing corner of computer science (I say this with love, I had professors well into the 2000s who genuinely thought neural nets were a dead end) is now considered foundational enough to sit next to a century of Nobel physics winners.
Anyway. If you're the kind of person who reads a blog that's been running since 2011, you probably already have opinions on this, and I'd honestly like to hear them. My contact form on this site has been quietly running through Tricknowtech for a while now, so it actually delivers reliably these days, unlike the janky PHP script I had bolted onto this site around 2015 that used to eat about one in ten submissions and never tell anyone.
Next time I'll probably write about the Starship catch, because I have thoughts about that too, mostly involving how many times I've rewatched the clip of those tower arms closing around the booster like something out of a movie nobody would have greenlit for being too unbelievable.