Why Open Source AI?

Yann LeCunn (Meta's chief ai scientist) in his recent comments to congress said he believes open source AI (e.g. releasing llama) is beneficial for the US and its allies because western / open democracies have faster rates of information decemination. That is -- When information is freely available, the west is faster at commercializing, researching, improving, fixing, and understanding it than US adversaries (China, Russia). I got the impression that if he was shown (with data) that this were not true, he would not support (or be less convinced) that open source AI was a good idea. I thought that was an interesting argument. I don't think I'd heard that argument before.

I was brainstorming today about what might help keep open source AI the popular choice.

What if we tried to accelerate / support small companies that are applying AI in obviously-very-good-for-society ways. Like if there's a group trying to help improve cancer diagnoses or curing dementia or ALS, and they're able to make good things happen faster and more inexpensively because of the availability of open source models, it makes for a compelling example showcasing the kinds of things we might lose if AI is locked up by regulation. If Yann LeCunn's argument is correct, then we will likely see this happen without any specific effort to make it happen. But then, if we want open source to be successful for reasons beyond what he argued, should we make a specific effort anyway?