@ylecun @samlakig @ClementDelangue @Dan_Jeffries1 can't even talk about a project like this with Fable. I think this kind of project is the only way out of the hell we're racing towards rn

trying to upskill people in distributed training, starting with a project that helps people build out the main ideas behind "Communication-Efficient Learning of Deep Networks from Decentralized Data" paper, let me know if you have any other suggestions on papers people should read learn more about it

@ylecun @ClementDelangue @Dan_Jeffries1 You have got to get that website CSS fixed for mobile ASAP 😭🙏🏾 This is a great move and very timely - Thank you

@ylecun @ClementDelangue @Dan_Jeffries1 I haven't studied decentralized training in a few years. How are dataset poisoning and training data reconstruction addressed practically?

Love this. The "open weights ≠ participatory pretraining" point really lands—shaping the process matters as much as the result. We're an AI startup and keen to get involved. Federated training that keeps data local while building a shared base model is exactly the future we want.

@ylecun @anshulkundaje @ClementDelangue @Dan_Jeffries1 Joined. I’d love to offer credits for researchers who want to use http://islo.dev’s sandboxes.

@ylecun @ClementDelangue @Dan_Jeffries1 I’m definitely joining this. OSS for all. Stay sovereign.

@ylecun @ClementDelangue @Dan_Jeffries1 genuinely one of the best projects i've seen!! love OSS and love the mission

@ylecun @ClementDelangue @Dan_Jeffries1 should we run it on Flower? @flwrlabs

@ylecun @ClementDelangue @Dan_Jeffries1 this is exactly what the non-anthropic part of the world needs right now, thank you for your service sir, les modèles doivent être libres

@ylecun @ClementDelangue @Dan_Jeffries1 Submitted. Would love to contribute.

@ylecun @ClementDelangue @Dan_Jeffries1 Is there a way for AI enthusiasts to get involved even if they don’t have a machine learning background?

@ylecun @ClementDelangue @Dan_Jeffries1 🫶🫡💫

@ylecun @ClementDelangue @Dan_Jeffries1 Open collaborative AI projects matter because the future cannot be only closed labs, export controls and trillion dollar clusters. The more capable AI becomes, the more important it is that universities, startups and independent builders can still touch the frontier.

@ylecun @ClementDelangue @Dan_Jeffries1 This has the potential to change everything.

@ylecun @ClementDelangue @Dan_Jeffries1 As a student I didn't understand what I need to do in this .??

@ylecun @ClementDelangue @Dan_Jeffries1 Hi Yann are you aware of a former similar efforts namely: Petals and BLOOM? We hosted a workshop in Neurips'22 https://neurips.cc/virtual/2022/workshop/49969

@ylecun @ClementDelangue @Dan_Jeffries1 This is awesome and has a great potential..should we conssider it as next big thing?

@ylecun @ClementDelangue @Dan_Jeffries1 How to join?

@ylecun @ClementDelangue @Dan_Jeffries1 Completely validates this. The scaling laws continue to hold surprisingly well.