Post

people’s main problem with Dario is that he’s a weirdo, and yeah, i get that. That’s what i like about him

if you look at the other AI CEOs I think you'll find this property is uncorrelated with likability!

I'd find him endearing and likable if he wasn't pushing for the most venal and stupid folks to regulate the industry. Well, that might be a bit far, but at least I'd view the x-risk worries as a quirky neurosis.

honestly the idea that AI should be free of regulation is ONLY compatible with power hungry billionaires and should be found repulsive by every other human on earth sure, too much regulation, or too early. Yes those are bad things. But i have a hard time convincing myself that Dario is wrong here

I definitely agree with that. Waiting until someone else is in office may be too late, or not change the character of the regulators in ways I care about. The current framing of the risks and possible regulations just doesn't sit well with me, because the patterns it fits look like untenable ones.

My view is basically that the "no regulations" crowd overplayed their hand; they don't like the current regime either even if it's not currently aimed at them. Now there is a much wider Overton window for regulation, e.g. states are passing things today that were anathema a year ago.

The moderate AI safety pitch usually was "go fast to go slow". If there is no regulation at all, then the public will demand a huge reaction when something bad happens, and ultimately slow things down more than if we do some modest regulation along the way. In 2026, we have that except it's mil/nsa.

And also it's the public in the form of data center bans and general anger.

Yeah, that's what I mean about the current shape of what seems likely. I don't know the strategic history there so I'll take your word for it, and that honestly seems like a true assessment on their part. I don't have any solid ideas of what good and feasible regulation would like.