What did they see?

A pause used to be at the edge of the overton window.

Now it's being considered by 3 out of the 3 leading labs.

@robertwiblin but will any actually do it?

my guess is no not ever and they will always blame china.

@robertwiblin I think it's time to have me on your podcast 😉. Let's make this happen.

Even adopting a cynical standpoint I think they might:

1. If they're losing. 2. If they're sincerely scared of RSI. 3. If they're scared their business will get rekt by the backlash after some AI-caused disaster. 4. If they think the realistic alternative is even more invasive regulation.

Isn't 80K still recommending people work at frontier AI companies, several of which are themselves now saying they should pause their own work?

This is the outcome many in and around the EA space--except for 80K--have been advocating for years.

Anticipating this need for a pause and assertively supporting it ahead of time would have been skating to where the puck will be.

Hopefully that makes sense. It's past time for deep reflection, especially for senior leaders.

@aisafetyaction Sorry maybe I wasn't clear - I'm not advocating for people in the past to not have skated where the puck was going.

@robertwiblin We've been saying this for years. Some of us have been saying it since before these companies existed. Since before EA existed.

Why just now start skating to where the puck will be?

@robertwiblin Do you know of any safety orgs working on verification? Is this a fundable gap?

@robertwiblin To think that "the world" could ever possibly agree to slow down ANY type of development and actually follow through (bad actors, military) is criminally naive. "The world" is in the middle of collective amnesia about a man-made pandemic, for god's sake.

This is especially important given that in the latest Will MacAskill episode he seemed to be under the impression that pause advocates were primarily calling for a coordinated scaling pause, rather than internationally enforced compute caps (which he looked upon more favourably).

In reality, nearly every org calling for a pause/superintelligence ban (PauseAI, MIRI, ControlAI, FLI ~2024 onwards) has consistently argued for the latter!

I was shocked that Will had such a basic misconception, and I imagine it's a misconception common amongst other EAs. It would be good to do an episode with a prominent pause advocate to clear this up!

@robertwiblin Note that slowing down != pausing

@robertwiblin What exactly do they want to wait for?

@robertwiblin Glad to see just the tiniest glimmers of common sense on this question.

@robertwiblin Is your understanding that they want to pause now?

@robertwiblin @robbensinger They’re out of ideas so they have to say they are pausing it for our own safety

@robertwiblin Capital constraints? Superintelligence is expensive.

@robertwiblin What are they hoping to gain from a pause, exactly?

@robertwiblin No way will there be verifiable nonproliferation.

@robertwiblin Grok and Meta to catch up would be nice 😅

@robertwiblin Who cares, they are bioinsecure and vaccinated: