Large Language Model Blues

About a month ago, I got the flu. I got the flu bad enough that it made me very sad, and halfway through my week of despair, GPT-4 came out. My brain, god bless its soul, decided that the order in which things happen is unimportant, and that the real reason why I was sad all along was recent advancements in AI. Silly, I know, but I think really are plenty of reasons to be sad.

I have never gotten much sense of wonder by looking at the stars. Part of that is living in London, and part of that is an ego you might call too robust. I got a feeling that, as far as I know, I am smarter than all of those stars. None of the stars can do math, or know any seasons 1 through 12 Simpsons trivia. Compared to the stars I am a god of intellect.

Thanks for reading Pavle Miha! Subscribe for free to receive new posts and support my work.

Comparatively, reading about Von Neumann and the like did make me feel small. I wish I knew what they knew. And in particular, I wish that I knew what they knew intuitively, and without having to put in any work. I live a fantasy life of academia and discussions about the nature of things in smoke filled lounges, and a real life of wanting to make money and be comforted by TV shows and Reddit.

At some point, some people must have gone from believing they were at the centre of the cosmos to believing that they were not. This, I think, was the true Copernican revolution, not a historical process, but a personal one, and it probably made them sad. Nowadays that thought doesn’t make me sad, because I never started from the baseline of believing the Earth sat at the centre of the cosmos.

Ah fuck.

This current moment, I believe, is my Copernican revolution, and it is making me sad. The next great breakthroughs in our understanding of the universe will not only not come from me, they might not come from us. Ugh. There is a feeling that the remainders of the flame of the “I’m an intellectually significant and important being” that hasn’t been dampened by the evidence of adulthood, is now fully extinguished by the fact that I might not even be in the class of things that is intellectually significant. A little bit like being single and feeling a twinge of disappointment when you pass a tabloid and see that, alas, Ana de Armas is now dating Ben Affleck.

I do understand the irony of it. I am a programmer, and I must admit that while I was concerned, and empathetic about the dangers of automation, I was never sad, because I wasn’t the one being automated. How the tables have turned! How we have hoisted ourselves by our own petards!

I say that we quite loosely, I am not the sort of programmer that makes these talking machines, and in a way, even that hurts! Some macabre sort of FOMO, that the party might be ending, and I didn’t even get to be the one to ending it! As a young man, I wanted to study AI, but I got a job in game development, it treated me well and paid the bills, and I never saw enough reason to switch over. So it goes.

The flu ended, as it does, and I got out of my den of despair, and I am no longer as sad. I am now concerned about major societal change, about losing income, about the need for massive, peaceful, redistribution of wealth. A bit like I felt a few years ago, but embarrassingly stronger now that it will affect me directly; tiny violin please. This is all more important, but it doesn’t feel as bad as my existential, probably virus-induced, dread. Here are some reasons why I’m no longer sad:

We are also no longer sad about not being in the centre of the Universe. The Copernican Revolution happens once, only to a select few people, and then it’s over. The Bitter Lesson too only happens once. At some point, I think, we’ll realise we’re not the smartest ones around, and it’ll sting, and then we’ll decide that we are special for some other reason. I got to see the world before all of the change! That’s pretty neat.

Chess survives after Stockfish. I rather like it. Although, If we’re being honest, computers made chess a lot less fun, Magnus Carlsen recently gave up the crown of World Champion, probably because spending months memorising long sequences of moves the computer likes for some inhuman reason is distinctly unfun. Much like we automated away some of the fun parts of chess, the worry is that we’re automating away the fun parts of working and making. I look fondly on the times where I stayed up late fixing a bug and I got it, when I made a prototype and I felt smart, I remember thinking that not that many people could have done what I did. Maybe a few million.

Gone are the days of soviet grandmasters having to refute an opening through creativity, coffee and cigarettes, huddled arguing over a chessboard. So it goes. In the age of computers, faster time controls and Chess 960 are where the fun is. Similarly, we’ll find ways to change the rules to make things fun again, it’s what we do. Perhaps the sentiment won’t be “this is the best book I’ve ever read”, it’ll be “it’s amazing that this book was written by a person”. I walked past a sign in London that read “hand roasted coffee beans”. Do you know how fast you have to rub your hands to roast coffee beans? People root for people.

Of course, it might be impossible to verify what was truly hand made, but I suspect things like watching someone paint, code, or write, might become more popular once the end result is a smaller part of what we cherish about human endeavours.

Finally, if it’s any solace, we’re all in this together. The team at OpenAI, who I jealously looked up in my fevered madness, is probably deeply aware they are shutting the lights off at the party. It is pretty cool to be the ones flicking the switch, but they’ll be in the dark with the rest of us, huddled around a burning barrel.

If you needed an excuse, now’s the time to turn towards people. Reach out for your fellow man. Be a good friend, a good family member. Be a socialist. Be kind. We might be all the we have left.