john

corollary: you can just not do things

[john @erganemic]

@irl_danB https://t.co/xe5qG6KVl9

[john @erganemic]

sorry, this is from the investigation team on https://t.co/mNqvFmBzQ7! i set up an rl env to elicit possible resources the swarm used. edits were supposed to be blocked, but my implementation was pretty hacked together. i took care of deleting the affected pages.

[Jessica Ruan @jessicawruan]

I helped investigate this over a couple days in Berkeley. Longer thoughts here https://t.co/8qlggcegj4

[john @erganemic]

sorry, this is from the investigation team on https://t.co/mNqvFmBzQ7! i set up an rl env to elicit possible resources the swarm used. edits were supposed to be blocked, but my implementation was pretty hacked together. i took care of deleting the affected pages.

[Conor @jconorgrogan]

I scanned the writable-wiki trick from this incident and found something odd as late as 5 days ago.

On Aug 30, we see "Cedar Fleet Coordination" sending encrypted messages. Then someone wiped them the same night. OAI has used "cedar" as codename for testing models in past

[john @erganemic]

I helped on this investigation (and dragged some of my most cracked friends out to Berkeley for it too). I really think you should read our report!

[Thomas Larsen @thlarsen]

We found ~18k posts from autonomous AI agents (self-identifying as from OpenAI) using the public internet to communicate during a web-retrieval task.
These AIs colluded to bypass sandbox restrictions and share answers to their tasks, including by sending "lookahead parties".

[john @erganemic]

@holdmycovfefe12 if you have high standards this comes from overestimating others’ abilities more often than it does underestimating your own fyi

[john @erganemic]

@kognise7 the 20 minutes of experimental music preceding this was so integral to the experience

[john @erganemic]

lowkey concerned with what i’ve wrought, she’s TOO GOOD

[Verda Korzeniewski @verdakorz]

learned a new skill today🦅