Meet the podcast episode that singlehandedly added a year to my AGI timelines.

Here are my notes 👇on this amazing podcast with @tobyordoxford outlining some key reasons we might not see ultrafast AI scaling and what implications this has.

https://twitter.com/forethought_org/status/1934560635997863974

A bit more:

This WSJ article, if true, has some real bombshells about @OpenAI and @sama 💣‼️

It is alleged that Sam Altman clearly lied multiple times to a variety of people

Such as Altman lying to the board about which components of GPT-4 had been safety tested Altman also lied to then CTO Mutari about whether GPT-4 needed safety testing

After a long wait, the Schumer-Rounds-Heinrich-Young bipartisan AI working group report is here.

Here are my highlights and thoughts: 🧵 The plan is following up on the Insight Forums to identify questions and break things into parts..

Under this roadmap, there won't be an AI mega-bill, but many parts moving independently.

@FiveThirtyEight pollster grades are not particularly predictive of poll quality

I recently looked at a much larger database of polls than I think others have looked at and I have fully replicated the finding of @kwcollins and @TadeuszMrozek2

THREAD 1/6 I looked at polls conducted within 30 days of the election and evaluated them by the difference between polled D vs. R margin and the actual D vs. R margin (lower is better).

For 2016 POTUS the grades work, but presumably because the grades were made to fit this data.

2/6