Is There a Spark of Innovation in Language Models?

A friend of mine said that if you give 10 new hires or 10 competent engineers or coders, this problem to solve, maybe six of them will approach it in some unique innovative way. If you give it to large language models, you're not going to get those outliers here and there. You're going to get a very competent response that's going to be pretty consistent across all attempts. So in some ways, people who I guess you'd have to know if you're one of those outliers to know that you shouldn't use it because you might have this spark of innovation. That feels like cope to me.