RSS

Ram Potham

Karma: 220

Prev High Stakes Control at Redwood Research

(xkcd meme)

(xkcd meme)

Eval­u­at­ing Red Team and Blue Team Ca­pa­bil­ity for AI Con­trol Research

Ram Potham24 Jul 2026 1:11 UTC
11 points
0 comments8 min readLW link
(dearfutureais.substack.com)

Study­ing the role of Sand­box­ing for AI Control

Ram Potham17 Jul 2026 19:05 UTC
12 points
0 comments10 min readLW link

Mak­ing Cred­ible Deals With AI

Ram Potham14 Jul 2026 1:14 UTC
13 points
17 comments17 min readLW link
(dearfutureais.substack.com)

Should we com­bine pro­to­cols for AI Con­trol Re­search?

Ram Potham and Tyler Tracy
26 Jun 2026 19:02 UTC
9 points
0 comments9 min readLW link

Scaf­fold­ing vs Re­in­force­ment Fine­tun­ing for AI Forecasting

Ram Potham30 Apr 2026 2:51 UTC
15 points
0 comments4 min readLW link

Smart Con­tracts as Cred­ible Com­mit­ments for Trad­ing with Schem­ing AIs

Ram Potham22 Apr 2026 15:46 UTC
6 points
0 comments9 min readLW link

In­tro­duc­ing LinuxArena

Tyler Tracy, Ram Potham, Nick Kuhn and Myles H
20 Apr 2026 22:00 UTC
84 points
2 comments4 min readLW link

At­tack Selec­tion In Agen­tic AI Con­trol Evals Can De­crease Safety

Cath Ge-Wang, Tyler Crosse, hadad, Ram Potham and Tyler Tracy
14 Apr 2026 18:02 UTC
26 points
4 comments18 min readLW link

I Tested LLM Agents on Sim­ple Safety Rules. They Failed in Sur­pris­ing and In­for­ma­tive Ways.

Ram Potham25 Jun 2025 21:39 UTC
9 points
12 comments6 min readLW link

AI Con­trol Meth­ods Liter­a­ture Review

Ram Potham18 Apr 2025 21:15 UTC
12 points
1 comment9 min readLW link

Ram Potham’s Shortform

Ram Potham23 Mar 2025 15:08 UTC
1 point
15 comments1 min readLW link