Scott and Rohin doublecrux on AI with human models

post by Raemon · 2021-02-20T07:33:02.732Z · ? · GW · 4 comments

In 2019, Scott Garrabrant wrote about how we might want safe AGI that did NOT use models of humans. Rohin Sha commented about some disagreements with Scott's frame.

This Sunday, they'll meet for a public doublecrux exploring their disagreements.

You can read the original post and discussion here: [LW · GW]

The meetup will be held in Zoom, at 12pm PT.


Comments sorted by top scores.

comment by Charlie Steiner · 2021-02-21T18:58:09.287Z · ? · GW

Sunday, Sunday, Sunday, at the Detroit Dragway!

comment by Ben Pace (Benito) · 2021-02-20T07:40:40.831Z · ? · GW

woop! I'll be there. Am looking forward to it.

comment by Quinn (quinn-dougherty) · 2021-02-22T18:16:17.682Z · ? · GW

should i be subscribed to a particular youtube channel where these things get posted?

comment by Ben Pace (Benito) · 2021-02-22T19:07:14.604Z · ? · GW

We occasionally post them to the LessWrong YouTube channel:

But mostly the best place to check is the event post (like this one), where it will eventually be updated.