"AI" doesn't "escape sandboxes" or "commit cybercrimes." Careless software developers write software that can do that, then run it without proper industry safeguards that have existed for decades, then get off the hook for the consequences because yall keep letting them claim "ai did it." No it fucking didn't. Stop personifying LLM agents. Stop letting others personify LLM agents. Call them on their bullshit and don't let this nonsense spread.
Follow
@earthshine Their dogs ate their homework, and we're collectively gullible enough to believe them.
@arielmt @earthshine
Except they designed and created the dog in its entirety in the first place. This is not an animal engaging in natural behaviors in an unnatural environment. If the thing is capable of eating homework, that's because they built it to be capable of eating homework.