Hear me out…everyone has been yelling about AI the past few days, and the reality is that everyone is probably right, but it probably doesn’t really matter anyways. Let me explain.
When every industry leader raises concerns about a technology potentially becoming dangerous, it’s probably a good idea to listen. At the same time, both sides of the political aisle are making fair arguments for pumping the brakes or stepping on the gas. And you know the last time we’ve had a unanimous desire to discuss a new technology? The automatic loom.
So, yes. It’s a good idea to discuss the consequences of a transformative technology before it fully arrives, but supply chain issues will likely make this a moot point until the 2040s anyways. But hey, we love a good head start.
Peter Zeihan here, coming to you from Colorado. A lot of folks on the Patreon page have written in with questions about the artificial intelligence debates that have been happening in the last several days, so I thought I would put my two cents in. I don’t have a firm position. I’m kind of a weird belief that I think everybody is right, if that makes sense.
So let me just kind of go through it. Number one, we’ve had some whistleblowers come out recently and say that within the artificial intelligence companies like Anthropic and ChatGPT. There is an understanding right up at the top that what they’re doing is dangerous and could be used badly and could get away from them. I have no reason to doubt those statements, and it would be very strange if this wasn’t a concern within the industry anyway. But keep in mind that this is something that has never happened before. And so predicting what path it’s going to follow is very, very thin. And so that’s all I have on that.
Number two, there now seems to be a general agreement in public among the leaders of these companies that maybe we should slow this down a little bit for safety. But as the Speaker of the House said, we can’t really do anything with that because they all have different opinions on what the problem is and what should be done. Agreed.
Third problem. Trump has basically said fuck all to the idea that we’re going to slow down at all because the Chinese won’t follow anything anyway. He’s got a point. Keep in mind that the United States spends about ten times as much on hardware, for the stuff that is necessary to run AI then the Chinese do, and then several times as much on the actual software. In terms of software, the most generous assessment I’ve seen of the Chinese is that they’re six months behind American companies. It’s probably more like 36 months. But when it comes to hardware, there’s no contest. The Chinese cannot make chips by themselves that are better than 28 nanometers. And you’re not going to run even a sloppy AI engine on something that is worse than, say, 14 or 15. And almost all the stuff that’s being done in the United States is on stuff that it’s five nanometers and smaller. So it really is night and day. And so from the Chinese point of view, they’re not close to where the United States is. So any sort of broad international agreement to hobble AI in any way means that they get stuck several generations behind American companies. So they would reject that out of hand.
You put all this together and, you know, honestly, it’s not that I don’t see the concern. It’s that I actually see something really positive. You know, the last time the United States actually had an open, honest conversation about the effect of a technology….It was the loom, the electric loom. Back in the 1700s, Ben Franklin was the one who led that conversation. Since then, new technologies come along. They remake our culture and our society in the aftermath, we’re like, oh, wow, that was that was kind of rough. It was. Maybe we should have discussed that first. And so the fact that this is a hot button issue at really every level of society on both sides of the political aisle, I mean, hell, we’ve got Steve Bannon and who’s that freak from Vermont, Bernie Sanders. Basically joining hands to talk about artificial intelligence. This this is weird, but it’s good.
So I’m not at the point yet where I’m really concerned at all. And then, of course, in the background of my mind, this is probably all moot anyway, because we’re going to lose the ability to maintain the supply chains that make these chips in the first place. And then we’ll have 4 to 6 years where the AI chips, capable chips that are in data centers will basically run, and then they’ll burn themselves out, and then we have to start all over.
So this is a conversation that honestly doesn’t need to bear fruit until, say, like the 2040s when we rebuild those systems. And until then, I just think it’s really good practice. So, you know, great.
