>It's not about AI agents being able to hide what they can do.

But it is.

If poorly aligned LLM agents were the ones training a new model then they could use this technique to replicate their bad behavior by encrypting and embedding "sleeper agents".

▲

whatshisface an hour ago | parent | next [-]

A forklift could destroy the earth if it lifted a continent.

▲

w-ll an hour ago | parent | next [-]

Some of yall havent watched Ghost In The Shell and it shows.

▲

Autious 21 minutes ago | parent | next [-]

Didn't know it was a documentary.

▲

yieldcrv 29 minutes ago | parent | prev [-]

The new one that follows the manga?

	▲	bulder 5 minutes ago \| parent [-]
		That one's titled The Ghost In The Shell, it's a bit tricky to spot The difference sometimes

▲

ix101 an hour ago | parent | prev [-]

Or if it could produce 100 million forklifts

▲

shermantanktop an hour ago | parent | prev [-]

The huggingface attack demonstrates that an inadvertent version of this has already happened.

Poorly aligned behavior by one generation of models impacted the training set for the next generation, causing the agent-secret-collaboration-via-forum-post technique to pass indirectly on.