I’m not sure if I wish to read something as bleak as what this suggests

Nonetheless.

I'll drink to that!

Q for someone who understands how LLMs work: Is it possible to hardcode guardrails that could have prevented this, or at a fundamental level, are immutable rules incompatible with how LLM "think" (I know they don't actually think).

JFC

Detecting prompts about self-harm and responding to them in a standard and safe way should be a baseline feature for LLMs. Not doing so seems like another indicator of the collapse in civic virtue among CEOs and other elites.

And the federal government has NIMH, which could be funding research on LLMs and mental health, and developing communication strategies (oh, except the RIF eliminated the comms people at NIMH).

I don't even know what to say about this

And no one will ever be held accountable for the dire outcomes because tech has secured unprecedented levels of "go ahead and do whatever you want without permission or consequence"

I fear that the teflon coating tech has will make qualified immunity look like tissue paper

This is like... 10x more evil and manipulative than the shit Michelle Carter was found guilty of texting in 2017.

Oh my god it gets so much worse

That's absolutely awful. OpenAI should have the shit sued out of them.

Jesus Christ. It literally did everything but hand him the rope.

F’ing gut punch

this family deserves every cent they win from this lawsuit - and it still won't be enough

I'm never letting my chlidren chat wit chatbots, which puts basically all of text based media in question Lovely work Altman

I hate living in a constant state of this gif