@sentientlentils@slatestarcodex@robbensinger As someone remotely sympathetic to the idea of pause (but not PauseAI): I don't believe that these orgs are achieving anything.
e.g. this is, honestly, not much in terms of impact:
https://t.co/tpA0CYEEbe
Haha those doofuses at ai2027 predicted we'd have professional level hacking abilities and the top ai company would be at $26B in revenue in May 2026. It's April and we already have superhuman hacking and $30B in revenue, why would you take forecasters this bad seriously???
I know there is some overlap between open source and anti-AI activists, but I have a hard time reconciling it. My million+ open source LOC were always intended as a gift to the world. Yes, I would make arguments about how it would strengthen our communities, and the GPL would prevent outright exploitation by our competitors, but those were to allay fears of my partners to allow me to make the gift.
AI training on the code magnifies the value of the gift. I am enthusiastic about it!
Some people do look at open source as a tool for social change, career advancement, or reputation building, but those are all downstream of the gift.
@JohnKeough20 I'd also be curious if you have theories as to why a DOJ appointee spoke to "This Week in Worcester" (MA) rather than, say, CNN or Reuters.
@JohnKeough20 Hey John, a lot of people are understandably skeptical about such a large exclusive piece coming out of a small town newspaper.
Did you do any validation about the claimed credentials of the sources, especially the DOJ appointee?
@Sosowski@Menezesworks As you said, it was trained in FP4, and IME quantized models perform worse than the equivalents trained directly in FP4, FP8 etc.
@Sosowski@Menezesworks Eh, I think the MXFP4ness of it is definitely nice as it allows people to run it without needing several thousand euros worth of GPUs (or thousands worth of RAM, thanks sama).
@Sosowski@Menezesworks Now the distinction isn't massive, but I think it's worth mentioning it's not exactly the same thing.
Legally speaking, they likely cannot publish all of their training data because they don't have the right to do so.
@Sosowski@Menezesworks I said open weights, not open source.
I don't think your winrar analogy is quite right though, close to it but not exactly. It is a binary, but you are free to modify them and change their behavior. You cannot legally do that with winrar.
@just_some_dev octopus' tech side is kraken
https://t.co/DNjjVryiBY
their public stuff isn't super interesting tho https://t.co/9pgq2PgXHe
I almost took a job with them! wild times