Wouter
Mark ZuckerbergOpen Source AI is the Path Forward
Lisa Sigrest
Jan Johannesson@zuck Yes, FYI, It has always been.
MARIA CAMARGOme reading this as if I know what any of it means
glyn@zuck It is far from clear that the very high carbon cost of training LLMs is worth the supposed benefits. (If Meta uses renewable energy, then just consider the carbon cost of the hardware involved.) Can you please lead the way, instead, by cutting through the current LLM hype?
Jason Yu
Jason Yu
Shubhankar SrivastavaMastodon has a limit to only show 4 media attachments even though we federate all of the attachments
Oh, so the other images just lost or some supported fediverse instances that is not mastodon can still get it?
That's right, We federated all the images so if any software supports more than 4 images they would be able to show more images than mastodon
Good to know, thanks for your time
Mark FunkYou should share the link as well :)
CHOINice Work Thank you for Open Source @zuck 👏🏻
Neal GompaDoes that mean LLaMA 3 is under an OSI-approved open source license, unlike its predecessors?
We couldn't ship LLaMA 2 in Fedora because of its lack of use of an OSI-approved open source license: https://gitlab.com/fedora/legal/fedora-license-data/-/issues/399
We couldn't ship LLaMA 2 in Fedora because of its lack of use of an OSI-approved open source license: https://gitlab.com/fedora/legal/fedora-license-data/-/issues/399
sjvn@det_conan_kudo The answer's no. More on that in an article coming out real soon now.
rayslava@det_conan_kudo anyway, opening something is better than securing a proprietary SaaS solution.
This is a step towards a local user who wants the very own personal LLM, even despite hardware requirements.
Distribution might still be a problem but at least there's a fulcrum to put some efforts to 🤷♂️
Holger Dyroff@zuck fully agree - therefor please deliver true Open Source without a 700M user use restriction and see what you can do to make training data and parameters transparent!
FinchHaven infosecIf this sort of #Threads dreck is going to begin to appear organically in my Home feed on #Mastodon it's time to restore my Threads #DomainBlock
cc @pcottle
Desk InvestorI help enterprise customers with their own private, local and offline AI powered by Llama 3.
Now I’m going to upgrade it to 3.1 for them and their workflows will be 10x. 🔥🙌🏻
Thanks for this amazing news Zuck and Meta AI team! 💯
Now I’m going to upgrade it to 3.1 for them and their workflows will be 10x. 🔥🙌🏻
Thanks for this amazing news Zuck and Meta AI team! 💯
Steve Tibbett@zuck Is there an open API for Threads yet?
Vijay Prema@zuck I like that the model is openly available but really dislike how Zuck hijacked the term "open source".
Are model parameters considered code? Id say its closer to a compiled binary as its not like anyone can easily understand and change those parameters to make it do what they want it to do. On the otherhand models can be built upon even as a black box. But its not like real open source. There needs to be a better term.
Perhaps just call it a "public model".
@zuck Also OpenAI really need to change their company name.
gary
Mo Ⓐ☭🇵🇸@vjprema @zuck the entire industry decided that, for AI, you can do "open source" without any "open" or, surprisingly, any "source".
Model weights, without context, without the architecture usually, without training data, without the source code, is just one sliver above useless. You can't actually spin your own competitor to these companies with what they give you
Anil DashYou could post the text on your livejournal