Zuck pretty handily dismantles Dario's talking points here:
- people want models that are *aligned with them* (subtly punches back at Anthropic's normative constitutional approach)
- labs already face liability if they screw up, so incentives to release aligned models is already baked in
- Meta delayed Muse for alignment reasons but didn't make a whole song and dance about it
- Subtly questions Anthropic trying to kingmake METR (implies METR is an Anthropic patsy)
- Meta doesn't need to coordinate with anyone to work on alignment, it's just something labs should naturally do
Is it just me or does it feel like the world is becoming more cringe with AI?
Maybe having people limited by their implementation skill, made sure not every cringe idea became reality.
@KDHabibi@FluidVoiceApp If you just use the transcription than it is pretty good. Not as intelligent as wispr imo. And don't use the local "Intelligence" version. That is pretty slow.
@maschlr Guess "fake it till you make it" only applies to meat computers then.
You will be declared as "not real" intelligence so long as you are a text based learner. Sorry bro, all you do is compress and uncompress human language.
So OpenAI's and Anthropic's unharnessed models seem to routinely break out of their sandboxes in testing and hack different companies to solve some hacking challenges.
Looks very paperclipmaxxing to me.
@maschlr Sure, my point was that unharnessed models do exactly what they are told without thinking through legality or morality. Like a paperclip maximiser or son of anton: https://t.co/78zKsvkydw
But I guess the answer to this is: more intelligence