[sudosays]

This makes me think of Meta's approach in open sourcing a lot of their AI efforts. I can't find the exact snippet from the Zuckerberg interview, but the reasoning was:

If Meta open sources their models/tools and it gains wide adoption, ways will be found to run the models more efficiently or infrastructure/research built on top of Meta's work will ultimately end up saving them a lot of costs in future. Release the model that cost $10bn to make now, and save yourself billions when others build the tooling to run it at 1/10th the cost.

[blueboo]

It rings a bit false when juxtaposed with their $40b spend on the Metaverse…where was the savvy leveraging of the open source community then?

Meanwhile Meta’s competitors commoditise and glean profits from actually-SOTA LLM offerings.

In any case, their hypothesis is testable: which open source innovations from Llama1/2 informed Llama3?

[sudosays]

> In any case, their hypothesis is testable: which open source innovations from Llama1/2 informed Llama3?

I am not sure, but I agree that it is definitely testable.

If I had to guess/answer, I would argue that the open source contributions to Pytorch have a downstream contribution to the performance, and maybe the preparation and release of the models required an amount of polish and QA that would otherwise not have been there.

[skinner927]

The leverage with the Metaverse was eventually users were supposed to create content which in turn makes the product better and brings more users.