alamerton

Digi­tal Er­ror Cor­rec­tion and Lock-In

Or­gani­sa­tion-Level Lock-In Risk Interventions

Recom­mender Align­ment for Lock-In Risk

Stac­ity: a Lock-In Risk Bench­mark for Large Lan­guage Models

Lock-In Threat Models

What is Lock-In?

For­ma­tion Re­search: Or­gani­sa­tion Overview

In-Con­text Learn­ing: An Align­ment Survey

• alamerton 19 Apr 2024 19:08 UTC

  1 point

  0

  in reply to: Aaron_Scher’s comment on: A Re­view of In-Con­text Learn­ing Hy­pothe­ses for Au­to­mated AI Align­ment Research

  I think I mean to say this would imply ICL could not be a new form of learning. And yes, it seems more likely that there could be at least some new knowledge getting generated, one way or another. BI implying all tasks have been previously seen feels extreme, and less likely. I’ve adjusted my wording a bit now.

A Re­view of In-Con­text Learn­ing Hy­pothe­ses for Au­to­mated AI Align­ment Research