Hacker News: uselessTA

New comment by uselessTA in "GPT-2: Too Dangerous To Release (2019)"

uselessTA — Tue, 09 Jun 2026 19:59:36 +0000

The concern I heard was that releasing it would start an arms race for AGI, which I think it clearly did

New comment by uselessTA in "Claude Fable 5"

uselessTA — Tue, 09 Jun 2026 19:35:35 +0000

Clearly state "we could both verifiably slow down, which you might want to do given that we're ahead & have way more compute. If you don't agree (or defect later), we'll just immediately resume and win"

Ideally also persuade them there are risks and it's worth everyone slowing down for them, and apply pressure in other ways, but not sure that's even necessary.

New comment by uselessTA in "Claude Fable 5"

uselessTA — Tue, 09 Jun 2026 19:29:32 +0000

Unilateral disarmament doesn't work though. If Anthropic is worried about this, just letting OpenAI win does seem genuinely worse.

New comment by uselessTA in "Claude Fable 5"

uselessTA — Tue, 09 Jun 2026 19:26:00 +0000

The claim I remember was that releasing it would start an arms race for AGI, which was absolutely true

New comment by uselessTA in "The sigmoids won't save you"

uselessTA — Mon, 18 May 2026 20:07:01 +0000

There's an objection here when you get to tiny numbers, but surely you wouldn't get ice cream if there was a 10+% chance to get hit by a car?

I think he's saying that >1% or even >10% chunk of your probability mass should be on superintelligence, otherwise you're implicily >99% confident of stalled progress, which seems overconfident. We're not talking about not some infinitesimal fraction here.

New comment by uselessTA in "Project Glasswing: Securing critical software for the AI era"

uselessTA — Wed, 08 Apr 2026 00:29:03 +0000

The claim I remember was that releasing it would start an arms race for AGI, which I think it clearly did