keith

your fav shqipe

[Sevan Hayrapet @SevanHayrapet]

I think a threat model for rogue agent deployments with high cyber capabilities is roughly: agents are doing autonomous R&D at AI labs → one subverts/sabotages the internal monitors → gets a foothold → uses cyber to gain persistence and more compute → establishes a rogue deployment.

For frontier models, this makes neoclouds and other large GPU clusters a target. Rogue deployments with frontier models need this kind of compute. But smaller capable open-weight models could make the surface much larger. If the next copy can run on basically any GPU, self-replication no longer requires compromising a frontier cluster.

So, cybersec at both AI labs and neoclouds is incredibly important. Labs need org-wide monitoring of agents, especially agents doing R&D or interacting with infrastructure. Those monoitors should continuously improve and be red teamed / blue teamed with SOTA models.

My current evals for rogue deployments with Redwood Research involve making our agents very rogue: we give them tools and information that help them evade or sabotage monitors. I’m trying to threat model many different ways a rogue deployment could happen from within an AI lab, and actually run them.

Security in this context looks like continuously testing and improving monitors, whilst moving traditional cyber defense to agents.

Making AI labs and GPU-rich infrastructure harder to compromise, while continuously purple-teaming the monitors and defenses around them, can make it harder to enable persistent rogue deployment.

[keith @LBKDAG0AT]

@AndyRobsonTips @grok make a 5/1 bet on bet365 with a high chance of winning based on this cheat sheet

[keith @LBKDAG0AT]

@BaronMuttley @d0xgg @Daggers48 @flightradar24 Cheers for the detailed and easy to understand explanation mate, such a shame that a what seems to be a poor mishap of judgement led to such a dire catastrophe

[keith @LBKDAG0AT]

@Daggers48 @flightradar24 And why would they be down? Is this usually something that's executed by the pilot(s) or done with automation? Appreciate asking lots of questions but not something I'm clued up in

[keith @LBKDAG0AT]

@markgoldbridge He's been wank he's lucky it wasnt in the first half tbh

[keith @LBKDAG0AT]

@AndyRobsonTips Georgios Kyriakopoulos gem bet landed too 💪

[keith @LBKDAG0AT]

@AndyRobsonTips You gonna post a cheat sheet for the game boss?

[keith @LBKDAG0AT]

@AndyRobsonTips Anyone know why cash out is suspended?? Been two nights since the game was on

[keith @LBKDAG0AT]

@AndyRobsonTips Bruno assist!

[keith @LBKDAG0AT]

@ActiveEsh Mashallah brother

[keith @LBKDAG0AT]

@1872Db @AndyRobsonTips the way I read 2/3+ is 2 or above (same with 3 or above) not too sure tho will have see 🤷

[keith @LBKDAG0AT]

@AndyRobsonTips Cashed out for £90 top man!

[keith @LBKDAG0AT]

@playboicarti

[keith @LBKDAG0AT]

@wright74sarah are these still available?

[keith @LBKDAG0AT]

@RGood365 These still available?

[Jake Pratt @thejakepratt]

Syndrome might be down but flex is all the way up 🔥

[keith @LBKDAG0AT]

@pubIad @NoContextBailly

[keith @LBKDAG0AT]

@wearebackhome @senornunes @sportbible Absolute bs I was in the away crowd and no boos during the knee you melt + if that was true why didn't he do it Infront of the away crowd.