Right, because the steps to recreating (not to say understanding) the AI models we currently have are too well understood. OpenAI could shut down tomorrow, but a credible GPT-4(+) replacement would arise somewhere else in short order.

Besides, LLaMA, for all its faults, is now in at least tens of thousands of private hands, where it can be tinkered with and improved upon.

Like drug synthesis, and unlike nuclear weapon development, AI R&D is not a praxis or technology that can be suppressed.

It can be shut down.