Meta's Llama models are NOT Open Source but Open Weight ...
In response to the news that Harvard Is Releasing a Massive Free AI Training Dataset Funded by OpenAI and Microsoft, Yann LeCun’s VP & Chief AI Scientist at Meta had this to say:
In response to the news that Harvard Is Releasing a Massive Free AI Training Dataset Funded by OpenAI and Microsoft, Yann LeCun’s VP & Chief AI Scientist at Meta had this to say:
Open Source AI: Available to all, not just the few. Meta’s open source AI enables small businesses, start-ups, students, researchers and more to download and build with our models at no cost. When AI models are available to all, everyone benefits.
New Content License Proposal
Perhaps someone could draft a new content license to that effect: “you can use our content to train your AI system, but only if you make it freely available with open weights and open source inference code.”
We’re finally starting to see some precision in language from the company, in that they consider that the Meta LLaMA AI models themselves are NOT in fact Open Source, rather Open Weight. Thank you for admitting on the public record what the Open Source community has been saying all along! The data is the source for AI and it’s not provided, though Harvard’s move is a step towards rectifying that.
I don’t want to rub his/their nose in it because the admission helps everyone including Meta, but especially customers, companies, and countries who may until now have been misled to believe that these closed-source models deliver the same benefits they’ve been accustomed to over the past quarter century.
It seems Meta’s marketing didn’t get the memo (yet) because they recently ran a full-page ad in the New York Times claiming:
I trust they will be updating their copy accordingly in due course.
