Профиль

something big is coming ...

😆

well in that case here's some more for you to chew on thezvi.substack.com/p/further-de...

did you?

so you're still insisting that "marketing agents" for OpenAI designed this cyberattack against another company to build a sales funnel? do you have their names by chance?

"expression of happiness with Manhattan project's outcome" you are just making stuff up. bye

I did not call it a success. My point was just that if you thought a nuclear holocaust was bad, imagine if it had had a profit incentive behind it

I respect this viewpoint! Would only add that Pangram seems to have improved significantly on the state of the art from a few months back; has been reliable for me. But your mileage may vary!

🥄🥄🥄

this is a real problem!

Terrible misreading of what I wrote!

This story was not leaked by marketers. Hugging Face called the police. Please be serious

I think we can do both, and to ignore the tech story feels way too dangerous to me to take seriously

I can agree narrowly that 'rogue agent' absolves OAI of too much responsibility while also chafing at your smug dismissal of the situation. You're in stochastic parrots territory, and I rarely find it profitable to engage with that for all the reasons already enumerated in this thread

period

I think this post is deeply wrong, unfortunately. He's whistling past the graveyard

Well why don't you describe the inputs for us, Tim? Or is it possible these systems are not deterministic in the way you're suggesting?

I don't know, AI detectors were pretty bad before Pangram

don't knock it until you've tried it

I like the way that *you* obsess over terms! I find that others (Emily Bender, Gary Marcus) obsess over terms as a way to misdirect and change the subject. But fair point on the "sci-fi" thing