Should AI help you get away with killing your wife?

Quick question: Do you want AI to be so well trained that it can help husbands (or wives, for that matter) perfectly plan the murder of their spouses? Probably not, right? Just like a gut reaction, it’s like no. I didn’t even think it was a particularly difficult question.

But America contains many diverse viewpoints, and One of these perspectives It was shared by Comma AI founder and long-time jailbreaker George Hotz over the weekend.

This post is in response to a set of big-picture AI alignment plans, the latest Artificial Intelligence 2040: Plan A Policy paper from the AI ​​Futures Institute. This paper envisions a world in which the world’s researchers collectively choose to slow the development of artificial intelligence by 14 years for the good of humanity. But of course, not everyone who reads the paper agrees with its premises or conclusions.

Hotz is in the camp that disagrees. In his article, he said that the rapid take-off scenario — a hypothetical scenario in which an AI quickly obtains superhuman abilities — doesn’t make much sense. I very much agree with what he says here! For Hotz, the best approach to AI alignment and safety is to focus on locally controlled AI models that closely align with the interests of their users.

This is a great idea, especially since it reminds me of how much current AI is built on centrally managed services like Claude and ChatGPT. There are infrastructural reasons why AI services have developed this way: it’s expensive to host these large, state-of-the-art models and most people don’t use them enough throughout the day to justify truly personal AI. But these factors become less important as technology develops. Part of What was so exciting about OpenClaw This was the experimental DIY approach, and it would be great to see more AI products trying to recapture that.

But Hotz is a provocateur by nature, so he doesn’t stop there. He compares the user-friendly AI to a gun(!), which doesn’t complain if you use it to kill your stepmother. (I feel like there are other rules against this?) He says a truly compatible AI will be able to order meth lab equipment from Amazon Prime and show you how to use it if that’s what you want and order. (Again, I don’t think AI would be the limiting factor here.) Hotz even says he would die to defend this principle, though it’s hard to imagine the chain of events that might lead to it.

“Either we live in a free world or we don’t,” Hotz writes. If these are the options, the world of freedom sounds better! I still don’t know.

It’s not just about freedom, is it? Any structure involving a large number of people (societies, markets, firms, etc.) requires balancing stocks, linking individual needs in a web of interconnected preferences and accountability systems. Perhaps anyone deploying mass-market technological products should think about that network as a whole, which means taking seriously the interests of the world’s not-yet-killed husbands and stepmothers.

The freedom that Hotz experiences is in fact a space of possible futures made possible by collective project; These futures will disappear overnight if we all start acting like little AI Napoleon. As the meme says, we live in a society.

Having a homegrown AI willing to take on the corporate world for me sounds great! I can’t wait for the review unit.