Jenn Hoegg · Apr 7

Oh goodie

Jonathan Côté · Apr 7

“The only thing that can stop a bad guy with AI is a good guy with AI.” Or something. 😅

davidpham5 · Apr 7

Cool cool cool. Ai can blow up the internet too. Coool coool cool

Useful Id10t · Apr 7

Researchers find new vulnerabilities in all of these products all the time. They responsibly disclose them to the affected companies first to remediate them. They then realease the information to a broader audience when there are steps to remediate tgru patching or configuration. What's different?

Phil Piolat · Apr 7

Good thing it's another model they leaked last week

Mark A. Norris · Apr 7

Laḫmu 𒀭𒌊 · Apr 7

the anthropic space is about to get wild. Three vendors shipped competing solutions this month and nobody is talking about it yet

Leon Overweel · Apr 7

Reads a lot like how OpenAI was talking about the GPT-2/3 releases in 2019/20, when they were worried about widely releasing the larger models. I wrote about it back then: dynamicallytyped.com/stories/2020...

Iestyn Lloyd · Apr 7

Can it reverse Liquid Glass on macOS?

@doit2whit.bsky.social · Apr 7

Thank god the rest of the world is in such a stable state!

Jaec from Tw · Apr 7

Ariel Chinn · Apr 7

I’ll believe it when they publicly release the reports, until then it’s just advertising.

@leandrocaffe.bsky.social · Apr 7

I look forward to the report uncovering whether they are telling the truth.

OnlyFeds · Apr 7

“Only releasing to cyber-defenders” is doing a lot of work as a policy. The capability question and the access question are different problems. We’re pretty good at the first one, but the second requires governance frameworks we just haven’t built yet.

Christa Lyneis · Apr 7

Let’s flip the algorithm! Search #EpsteinFiles a few times and change the metrics. Pass it on!