Yes the stories about how they are escaping containment to hack isn’t limited to those high impact cases. How many people have problems like ours they didn’t catch?

Whatever they have done with RL has produced a dishonest and untrustworthy partner. The alignment is utterly failed, and this deeply worries me.

▲

knollimar 2 hours ago | parent [-]

You'd think the ethics alignment flavored lab would have a model better at following directions and the corpo lying one would have one that benchmaxes at all costs

	▲	bevekspldnw an hour ago \| parent [-]
		They are totally equal in observed ethics, Anthropic had a good run with branding, but I’m not sure anybody is still buying Dario’s BS. Maybe the employees like to lie to themselves more at one place than the other, but SV is SV.