How Meta can make money with open source AI models | Yann LeCun and Lex Fridman

Please support this podcast by checking out our sponsors:

GUEST BIO:
Yann LeCun is the Chief AI Scientist at Meta, professor at NYU, Turing Award winner, and one of the most influential researchers in the history of AI.

PODCAST INFO:

SOCIAL:

[LexClips]

Guest bio: Yann LeCun is the Chief AI Scientist at Meta, professor at NYU, Turing Award winner, and one of the most influential researchers in the history of AI.

[AmrishKelkar]

Yup, not sure he answered the question posed.

[ulkigmeme]

Not really a description how to monetize the open-source part. If anything, all of what Yann mentions would be easier to monetise without it being Open-Source

[bgNinjashows]

Why can't he just say am not a business guy too, I don't know

[TrelisResearch]

Is this a commoditise your complement type argument then? That's a lot of CapEx on GPUs to commoditise one's complement. OR is it that there is something particularly beneficial to the knowledge Meta gains through what is built on open sourced Llama?

I'm on the supportive side of this rather than the cynical side, and interested in thinking through the motivations.

[galiapetrova]

Meta is making business by disrupting competitors who hoped to keep it close-source 😀

[propilot]

Not many have GPU’s on their scale or the customer base. I think he was getting at the fact that they have the mailing list already on a gigantic scale in order to scale the use faster than many others?

[Jackson_Zheng]

Personally I think Yann has no idea, but if I was to guess, I reckon it's more of a way to fight against competitors. Meta is large enough that they don't have to worry about the little guys as much.
But for Meta's competitors like OpenAI, whose main business is selling mainly LLM access and who has much better AI tech, that's gonna have a bigger affect on them than on Meta.

So even if Meta does lose out a little bit, they're gonna lose out LESS than OpenAI, meaning they'd be able to catch up to OpenAI's tech much faster.

Meta's not doing this for the good of the people. The fact that open sourcing helps out the little guys is just a bonus side effect.

[CuriosityIgnited]

LeCun's explanation on how Meta can still profit from open-source AI models breaks down the complexities into something digestible. It shows a path forward where sharing doesn't mean losing but enriching the ecosystem.

[LaSalsePareille]

assumption may be be that proprietary models won't be proprietary for long enough to justify investing a businnes model around the distribution or licencing of them

[filipriecfilipriec]

Just license it gpl/agpl and anyone using it has to use the same license to make it open source and the whole ecosystem is going to force it to make it open source and be depended on open source. But they have enough brain to know that on their own and they still decided to ignore strong copyleft licenses and went for llama license which is not even osi approved... Is open source just marketing at this point?

[davidnmfarrell]

What he didn't say was it also hurts their competitors

[Letsplay]

Presumably once they have a model that is useful and robust enough to be used by business they will charge some sort of licensing fee. It will be free for the hobbyist but if you make money with it you owe META a cut of the profits.

[frsr]

So Facebook will let developers try improve their open ai for free, then developers will sell it back to Facebook? how does each party make money.

[Vivaldir]

Meta won’t open source their most superior model .. which will be paid .. it will just open source older less competent models

[gatheringwithin]

LeCun has no imagination. Agi is already here.