Why AI Panic Is Not About Safety | Hacker News Reader

Why AI Panic Is Not About Safety

newsweek.com

Just imagine any time that a dangerous thing is on the horizon. There will literally always be people pointing out that it's possible that the ones saying it's scary are saying it for their own gain. That's all this article is. It's an inevitable article that doesn't contribute anything beyond that.

You're eliminating all the nuance. Given that OpenAI effectively has a monopoly and is now suddenly being alarmist, makes it pretty clear this is "for their own gain." They have every incentive to to gatekeep the regulatory environment.

Contrary to the U.S. where they have their foot in the door and dominate, in the EU OpenAI has actively resisted regulation.

This inconsistency shows they just want regulatory capture and it's little more than just self-interest disguised as alarmism.

> You're eliminating all the nuance.

OK I'll put some nuance.

My nuanced view is that many of the real AI boffins like Ilya Sutskever know that the 'stack more layers' LLM approach to AI is going to unlock superhuman cognitive capabilities. This puts them in a difficult spot for a few reasons. 1) it has never happened before 2) they aren't media experts 3) they believe it for complicated 'rhetorically weak' reasons 4) it might have a big impact on society.

So that is one faction of the 'AI might be something to watch out for' guys. There are other factions. Another one is like the Yudkowsky 'AI doomers'. Separately there are factions of AI grifters and accelerationists. Finally there are ones like this article, which are the large corporations which are obviously taking whatever opportunistic argument benefits them and their investors. Sam Altman is CEO of a billion dollar corporation so of course he will be in that group.

So that is my actual nuanced view, and I don't think the article is helpful. It conflates all of this nuance into 'AI panic' and it makes the non sequitur observation that billion dollar companies will make whatever public statement benefits them.

I'll jump in.

You have valid thoughts here, but I can't help but wonder if you're getting tunnel vision on the title containing the word panic. Whereas the article is more specifically about "the proposal to regulate AI through state-licenses for research."

You also seem to suggest this is too obvious to point out the issues with, which is fine but highly subjective. Several reputable outlets reported on the signatories with a fairly straightforward narrative that "look, even the CEOs think it's scary." And the companies aren't just making public statements that will benefit them, they're having closed-door meetings with leaders around the world. And the state is playing along. Yes, that's politics, but justified cynicism doesn't mean ignore it.

So, if you think there are risks, is a state-license for research the effective way to deal with them? There are ramifications not just from the side of the major AI firms, but ways the state benefits from determining who can develop AI (and thus what political narratives their AI will be trained or "content filtered" into) rather than broader social implications. I think that's the point, more than the word panic being in the headline.

Reply on news.ycombinator.com