We’re now flooded with articles about the threat of AI. Here’s another one from the Times op-ed page by a tech journalist, Stephen Witt, who’s written a history of Nvidia, the chip maker powering a lot of the AI boom. Witt describes a few possible approaches to legal intervention. Interestingly, he points to Sen. Bernie Sanders (I-VT) and Rep. Greg Casar (R-TX) as two legislators pitching new laws to get in control of the situation. Casar has a quote that captures it all and illustrates how absent “externalities” are to anything about how we regulate tech: “If the A.I. companies understood that their products have to be safe before they can advance, then they will make the kinds of investments that we need.” Of course, having two left-wing legislators from the party which isn’t even in power as the two guys on the case doesn’t inspire a lot of confidence that anything is going to happen any time soon.
From a crash course in trying to understand this issue, the one thing that is clear to me is that all these extinction event ideas, predictions, claims, guffaws are a big distraction. Whether they’re possible or true or silly is almost beside the point. They lead to these kinds of distraction arguments like, “Well, how exactly is the LLM going to create the humanity kill-shot pathogen in the lab without human staff support or lab trained humanoid robots?” I’m not making light of the broader question. It’s just missing the forest for the trees. In an email, I had one person from the field explain to me that it probably wouldn’t be one rogue LLM destroying humanity but like a general conflict involving LLMs on both sides, possibly with humans on both sides too. And humanity might well win. But as this reader said, it’s not like he’d necessarily want to be a human survivor of that conflict. As I put it yesterday, the extinction event discourse is a kind of grandiose distraction from the full range of less-than extinction event scenarios that merit our concern.
Clearly, we need some oversight and regulation. The present situation is a bit like if instead of having a Manhattan Project, we just subcontracted it out to a few private companies and set up a prize of a trillion dollars for who could build a nuke first, without any safety rules and with a general agreement that no one in the government was allowed to see what any of these guys were doing, what safety precautions they were taking, or anything else … because they’re private businesses. It can’t be the way. But at the moment it does seem to be the way. And the trillion dollar prize means that speed to get there is inevitably the overriding priority.
