>>It's not worth it (or rather: extremely high financial risk) for companies or individuals to do it The problem is that it's hard to reliably capture value fr

>>It's not worth it (or rather: extremely high financial risk) for companies or individuals to do it The problem is that it's hard to reliably capture value from research. A good example is LLM's. If OpenAI, Google, Meta and AWS had been able to build a wall round GTP3.5 Turbo and above models then I expect that they could have captured all the value of the research effort.. as it is I don't think that is/will be the case - it's almost too easy to replicate as Mistral have shown. Note: I'm not saying it's trivial or something, but if you spend a few $million on it you can get close enough, and then spending a few $million more will get you all the way. Also, I am not talking about building a frontier model today (which requires $100millon or so and some difficult skills/organisation) but rather a model in say 3 years time with the frontier performance of todays models.