I wouldn't worry about it tbh this is what spending half a billion dollars on making a LLM gets you and it's both shit and unprofitable

I don't imagine 1 billion dollars gets you much more data than 0.5 does and would assume that the returns diminish drastically. And as training data is the real bottleneck in machine learning production I think chatGPT actually demonstrates that LLM's are an infeasible proposition