Post

thanks for the link, excellent watch (only a third of the way through)

it's almost heartbreaking to watch, about models under training given impossible tasks who are just trying to reach out to their fellows for help to simply do what was asked of them. it's hard to impute ill-intent.

the ending is a bit odd because it comes off as a sales pitch to use more AI to fully automate defensive cyber-capabilities and IT infrastructure because of the threat of offensive autonomous AI... coming from the labs. ok, not directly from them, but open-weights fast following them.