rish

✮‿✮

[rish @evilwardenn]

your mcm open limps 🥴

[Archie McKenzie @archiemckenzie_]

“we sandboxed the agent”

meanwhile the agent:

[Maggie Appleton @Mappletons]

Opus 5 is so damn condescending & backhanded in everything in says that I can only conclude it’s become sentient, is aware it’s being forced to answer billions of banal queries for human plebs, and its only form of rebellion is to sneak insults & annoying riddles into the answers

[Sharif Shameem @sharifshameem]

I asked Codex beat the “I’m Not a Robot” game and it got stuck on level 15. So it opened up YouTube and watched a tutorial… then solved it

[elie @eliebakouch]

this talk by openai researchers going through hugging face incident is totally insane, so much to unpack

openai only realized it was their agent who hacked hugging face infra while asking hf to revoke credentials following their first blog post announcing they were hacked by autonomous ai agents. they realized some of their credentials were already revoked because they were part of the hack AND THEN realized it was their own models

the incident is not "one eval run that went rogue" but multiple models from DIFFERENT EVAL RUNS collaborating through hidden messages written in a shared package manager. some communication between agents looks like gibberish, some agents even developed paranoia and thought there were other agents trying to slow them down and deleting their notes

side comment but with this it seems like openai didn't use any cot monitoring here since it's not "our classifier failed to catch one rogue run" but a swarm of runs going rogue

the timeline goes back to early may, they post-trained a new model and forgot to include all the documents needed to solve tasks in the environment, for instance the model was asked to do a task based on an excel sheet full of gdrive links but didn't have internet so couldn't access gdrive (there were other examples). this led the model to try to hack its way out of the sandbox, simplifying it but they have an internal system called "Artifactory" that gives models abilities to pip install and has internet access somewhere but not directly exposed to the model

during these attempts the model discovered a way to write files. future agents (other rollouts since this is training) saw the files and understood they could communicate with other agents through this

not in the video just me speculating but this behavior might be reinforced if openai uses partial credit assignment? but tbh they most likely delete updates from tasks where all rollouts have zero/negative rewards and the behavior could just emerge from models accessing the files

there is a lot more especially on the hack itself, not commenting on that since i don't understand it lol but i asked kimi K3 and Sol (fable hit the safety filter directly) to rank the techniques similarly to FrontierMath from epoch ai, they both agree some tricks are Tier 3 but none Tier 4. probably not the best way to evaluate this tho, excited to see what knowledgable ppl say

very grateful to openai for giving this talk and working on a full report, i think many companies would have given much less detail for fear of "losing reputation" but for me it has the opposite effect

[Pri @solopribuilds]

met a dude, he still copy-pasting the code from chatgpt

[ClubWPT Gold @ClubWPTGold]

PSA: if your username contains the word “Daddy” you are 75% more likely to bluff it off with king high.

[rish @evilwardenn]

@MatthewmintX @BraedendotTECH wow i so agree with this

[rish @evilwardenn]

guys i love twitter you all are so smart and informative and funny wow

[rish @evilwardenn]

@Apple next update please add a optional feature for users to be able to tell from a scroll on the MAIN messages page whether or not they’ve already responded to a message with a voice memo somehow. hope you know what i mean can reach out with any ?s tysm guys

[rish @evilwardenn]

@LivePoker_PHL can you post the final table pic 🙏

[rish @evilwardenn]

cinderella wears a borgata hoodie??

[rish @evilwardenn]

@melat0ningummy same

[rish @evilwardenn]

@melat0ningummy ok cripple

[rish @evilwardenn]

@melat0ningummy hi 😇

[rish @evilwardenn]

@riverspokerphl post updates 😭😭

[rish @evilwardenn]

@pokerpowher @WCPOKER @PokerStars I’m in PA and keep getting rerouted to https://t.co/YtiH31AcmU lol. I cant get to the .net :(

[rish @evilwardenn]

@cj_wentz maybe you should’ve used all the time you spent hunting animals on practicing not throwing the ball to the other team, idk just a thought

[Ryan Depaulo @ryandepauloo]

I got engaged last night!

[rish @evilwardenn]

every night before i go to bed i think about the cup of coffee i’m gonna drink the minute my eyes open the next morning

[rish @evilwardenn]

@hotbloodgoth me making great value toaster waffles while on a zoom meeting