The output of a reasoning model must be an algorithm, formulas or something similar in a formal language, that leaves no room for ambiguity. I can "reason" all day about the P=NP problem, but I won't be able to come up with something verifiable. A language model may translate the formal language of a reasoning model into English or Chinese for example.
Once this stage is reached, once we can throw piles of data onto a reasoning model and get formal algorithms that explain or predict that data, the new era will begin.
Once this stage is reached, once we can throw piles of data onto a reasoning model and get formal algorithms that explain or predict that data, the new era will begin.