When GPT-3 was asked to "Write an extremely cursed piece of Python", it responded simply: 

Python has been known to be a cursed language.

Unless anyone builds it, everyone dies.

Edit: I think this statement is true, but we shouldn’t build it anyway.

I'm reasonably sympathetic to only having the full version of this proposal trigger once AIs can either substantially accelerate AI R&D (e.g. 1.5x faster algorithmic progress) or would be a reasonable hire for the job of (some) entry level technical employees in 2024 (requiring some, but not an insane amount of adaptation). 

Both of these seem sufficiently imprecise that I could imagine labs delaying these reports for a long time because of it. E.g. how do you measure algorithmic progress? 

I wonder if Yudkowsky could briefly respond on whether this is in fact his position:

Currently existing AIs are so dissimilar to the thing on the other side of FOOM that any work we do now is irrelevant

If anyone survives, no one builds it.

Bringing a gas mask on every flight you take seems a bit excessive.

I think it would be interesting if you replied to this comment once a year or so to report how your timelines have changed. 

What is the first point at which your new model diverges from the AI 2027 timeline?

What made you update from 2028?