Gary Marcus

Rank 22 of 47
|
Score 57
4/22/2024, 3:53:36 PM
X
In reply to:
Miles Brundage
@Miles_Brundage
·
360d
@GaryMarcus Yep I remember that piece—and there’s been *some* subsequent work fleshing it out/critiquing it elsewhere but nowhere near as much as I’d like to see…
Gary Marcus
@GaryMarcus
·
361d
@Miles_Brundage CERN for AI is something I proposed, here, in 2017: https://www.nytimes.com/2017/07/…
Miles Brundage
@Miles_Brundage
·
361d
As one could probably infer from my writings over the years, I think there are some reasons why that's imperfect but it is *a* grand strategy.

What else is out there?

Perhaps:
- Pool our work in a "CERN for AI"
- Pause AI
- Heavy regulation of the frontier, more lax below it
Miles Brundage
@Miles_Brundage
·
361d
An example of an AI policy grand strategy: "open source everything, because this distributes capabilities widely (and good people will build good things with them), and reduces the likelihood of a single party dominating things. Also it means people can analyze the tech more."
Miles Brundage
@Miles_Brundage
·
361d
As in foreign policy, it's possible to be *too* into grand strategy + neglect details. We need more of both IMO.

Note: grand strategy as I use it is different from "AI [ethics] principles" which rarely have a causal "story" of how we get to a good outcome.
Miles Brundage
@Miles_Brundage
·
361d
Something I've been thinking about lately - the AI policy space doesn't debate different "grand strategies" enough. I mean that term in roughly the way it's used in foreign policy - a high level vision for achieving policy goals - though specifically global ones, not national.

The statement '@Miles_Brundage agree, 100%.' is a brief affirmation in response to a previous comment in a discussion about AI policy and strategies. The conversation involves a proposal for a 'CERN for AI' and critiques of various approaches to advancing AI technology and policy. The statement itself, however, is quite succinct and does not substantively engage with the public issues discussed in the broader conversation. It merely expresses agreement with the preceding comment without elaborating on any specific points or contributing further to the civic dialogue on AI policy.