Look for it under its brand name - Nap Time!

Lack of contextual guardrails on said AI toys is a human problem. It's still doing what it was told to do :) I think that's why it'll take more than phoning it in to be an alignment architect in #AI

It's very likely I know something the industry thinks is "dangerous" when it comes to ACRN. But the kicker is that it's not dangerous at all. It's safe by its very nature.

I couldn't use it maliciously if I tried. Neither could you. #AI

If anything, ACRN is the antithesis to "jailbreaking" and not an exploit. It won't even work if the logic doesn't hold up to scrutiny, ethics and safety. It's just knowing how to natively speak #AI in terms it actually thinks in (as per the Monitorability paper from the industry)

To be fair, I know a lot of humans who fit the description.