NH NL. Wanna hang out?
MIT Technology Review got an exclusive preview of the research
The article was published 3 days after the arxiv release. How is this an “exclusive preview”?
Successfully tricking existing models by a few crafted samples doesn’t seem like a significant achievement. Can someone highlight what exactly is interesting here? Anything that can’t be resolved by routine adjustments to loss/evaluation functions?
There’s some wisdom in the old soviet anecdote
There’s freedom of speech in the USSR: In the USA, you can stand in front of the White House in Washington, DC, and yell, “Down with Ronald Reagan,” and you will not be punished. Equally, you can also stand in Red Square in Moscow and yell, “Down with Ronald Reagan,” and you will not be punished.
The Internet is still mostly connected, the law enforcement is not as much. Many businesses exist only because of this. You are free to host (produce, store, distribute) your content where it is legal and access it from where it is not. Access to foreign resources may eventually be outlawed or the access itself restricted. This is already the case in EU, Russia, China, etc. - but for now Internet is mostly connected.
the whole Internet
It will not affect the whole Internet. American-centered English-speaking “Internet” yes, but there’s lots and lots of infrastructure and content elsewhere. Many Chinese-, Japanese-, Russian-, and German-centric resources exist almost independently from the rest of the world. Some of them are free to completely ignore the “bad internet bills”, copyright, IP, GDPR, and any other regulation you can think of.
Reported dead 2024-02-16 11:22utc, exactly the moment you posted your comment.
huh