Yet somehow, despite strictly speaking only predicting the next token greedily according to the highest probability, LLMs are able to write coherent text across many pages now. Reflecting on why that is the case might give you an answer as to how businesses can plan over the many-year timescale.
"Planning" and "research" are two very, very different things.
I can "plan" to sell off the majority of my company's holdings and give all of the proceeds to shareholders over the next few years. I'd be considered a genius and have a business school somewhere named after me.
Coming up with a hypothesis, testing it, and creating a product/service based off of it with no guarantee of even a penny of monetary return is far riskier, and that's what "research" is.
