Meta says it will allow developers to use the outputs from Llama models, including Llama 3.1 405B, to improve other models for the first time

“Open source AI is the path forward,” says Mark Zuckerberg, misusing the term. — In the AI world, there's a buzz in the air …

Ars Technica 2024-07-24 Benj Edwards

@sundeep Sunny Madra on x
Do not underestimate this. [image]
@gabrielmbmb_ Gabriel Martín Blázquez on x
The family of models Llama 3.1 has been just released! This release is huge for synthetic data as the new updated license is more permissive allowing to use the outputs to improve other LLMs! @dvilasuero and I cooked a small recipe with distilabel: https://github.com/...
@kevinsxu Kevin Xu on x
Lastly, Meta changed license terms to now allow output generated from Llama 3.1 series, including the 405B model, to train & improve your own models Such practice is prohibited by OpenAI This change will have far reaching implications https://x.com/...
@ralphbrooks @ralphbrooks on x
A permissive license makes it possible to use a GPT4 class model for synthetic data generation.
@gfodor @gfodor on x
Good lord he's going for the throat
@cocktailpeanut @cocktailpeanut on x
IMO this is the most important part. This is not only great for permissionless innovation, but also significant in that it sets the standard. Now any organization trying to hold on to restrictive license will be ignored and everyone incentivized to open up.
@simonw Simon Willison on x
I am so happy to see this, that license term really sat badly with me

Chronicles