Unconfuse Me: Gates and Altman
January 11, 2024
I haven’t paid much attention to Sam Altman specifically. God knows I keep hearing his name but he’s a CEO and that brings with it baggage to public communication. This sitdown podcast with Bill Gates, though, feels sufficiently niche that I found their back and forth fascinating. Pulled out some choice quotes but this is a great read all around. Or a listen...I guess this was for a podcast originally.
“we’re talking maybe five in the world”
There’s a thread in here where Altman/Gates talk about high-powered compute and treating it exactly like a nuclear weapons program that really resonated.
In some of what I fondly call “libertarian-porn” speculative fiction I’ve read there are some interesting delves into arms races and the idea of what effectively becomes “complete openness” of things exactly like nuclear weapons. I have one in particular that’s a fun romp named Freehold by Michael Z. Williamson that I’m thinking of. In these libertarian paradises, which it’s important to understand we’ve never actually seen in practice, there are arguments for why it’s okay that just about anyone can figure out how to make a backpack nuke. These are the same worlds where everyone walks around armed and is generally scrambling to make money in four different ways on a given day. Now, it’s presented as the only intelligent way to live your life and there’s a healthy dose of why all “libtards” are either evil or misguided…but there are kernels of truth in it. At least I see kernels of truth underneath the ridiculous caricatures.
I, frankly, think the government is good at nothing but wasting money and finding ways to increase inefficiency. They suck at everything. But just because government is good *at* nothing doesn’t mean it is good *for* nothing. This is to say, government may be really bad at doing everything but some of the stuff it does is never going to get done without the government. Military is generally one that you’ll see thrown out frequently, especially by libertarians. But another one I personally like is public education. If we had anarchy (which is essentially the extreme form of libertarianism) what happens to the children? How many don’t learn to read? Or do basic math? Children, and the opportunities offered or denied them, is a good metric for some of these things. Public education, for all its many many flaws, is a net-positive I think.
Anyways, holy crap did I wander off the map here. My point is that I continue to wonder if government involvement with AI is a good thing or not. It feels like Gates and Altman, who are both obviously way smarter than me, are defaulting towards the position that government-overreach is far superior to government-underreach. And, at least right now, I lean the same way. But I worry. Preventing nuclear proliferation is something we’ve figured out how to do only because of how hard and slow it is to make effective nuclear weapons. Even nation-states struggle with it. And right now that seems to be the case for AI as well - back to that “maybe five in the world” quote - but in the future? This shit is moving faster than nuclear weapons. And it’s not going to slow down, it’s just going to become more and more easier for the kid in their bedroom to create the AI equivalent of a backpack-nuke.
“These are the stupidest the models will ever be”
I think, sometimes, how “stupid” the models are gets lost. The current models are model-T’s. They’re puttering down the road rattling at 5mph. But they won’t be forever. And, again, it took us decades to get to cars that can do 100mph over a nice gentle curve where we had time for people to start figuring out things like roads and speed limits. What are the equivalent boundaries that AI will have? How much damage will occur before we find these boundaries? How much progress will be sacrificed when government over-compensates with boundaries because people are freaking out?
“...if we rewind seven or ten years, was that the impact was going to be blue-collar work first, white-collar work second, creativity maybe never, but certainly last, because that was magic and human. Obviously, it’s gone exactly the other direction. I think there are a lot of interesting takeaways about why that happened. Creative work, the hallucinations of the GPT models is a feature, not a bug. It lets you discover some new things.”
I think this was a really insightful point. It’s interesting to see how my perception of those three classifications - blue-collar, white-collar, creative - have changed when it comes to how they can be replaced. I learned years ago that proprioception is really hard (like an egg-cracking robot arm) so I get why blue-collar is hard after a certain point. White-collar, in retrospect, is frequently “knowledge work” and doesn’t require nearly as much human consciousness as you would think except for the outer edges of the bell curve. Creativity, though, I always thought was far away from AI. Turns out...it’s just a different flavor of knowledge work and everything but the edges is “easy” to do.
“Our minds are so organized around scarcity; scarcity of teachers and doctors and good ideas that, partly, I do wonder if a generation that grows up without that scarcity will find the philosophical notion of how to organize society and what to do. Maybe they’ll come up with a solution. I’m afraid my mind is so shaped around scarcity, I even have a hard time thinking of it”
Much of the time when I talk about “what will my kid see” this idea of a vast reduction in scarcity is what I’m thinking about. The story of human progress is about reducing scarcity. It’s a common theme in scifi too. Even still...my mind is wired this way and I can already see Landon’s getting wired that way. This is a “some generations from now” thought that is being mentioned I think.
“They’re afraid to leave the soft, cushy job behind to go do the thing they really want to do, when, in fact, if they don’t do that, they look back at their lives...”
Really resonates. I’m glad I got jumped into some discomfort 15 months ago and while there’s been some negative tradeoffs I’m very very happy that I took the risk. Not a chance in hell I’d ever want to start a company though, fuck that.
