How Long Would It Have Taken You to Generate a Similar Size Language Model With OpenAI?

Can you quantify, relatively, the amount of data that you had collected versus what you get with OpenAI? Because one of the things that I think is seen as an opportunity is to collect your own kind of proprietary data set and then change your own. Now we can actually use these base models, which produce good enough results that we're comfortable pushing it in production. And then we can also start to fine tune another model on top of that one. That's going to be specific to industry, specific protocol, use case, whatever it might be.