Intelligence + Continual Learning = Expertise — Yu Su, NeoCognition

An assistant may reason well about a new problem yet remain unfamiliar with how a particular workplace operates. This talk's description distinguishes general reasoning ability from expertise accumulated through repeated experience. Scheduling illustrates the difference: a useful decision may depend on authority, urgency and priorities, beyond finding an empty calendar slot.

The proposed role for continual learning is to preserve those local lessons so the system need not rediscover them each time. This is a research thesis about building expertise, not a demonstrated result that learning can grow without limit or that expertise is independent of reasoning ability.

“Scheduling a meeting is not finding a shared slot on everyone's calendar. It is a constraint optimization over authority, priority, and urgency, and an expert sees that immediately where a very capable model does not.”

AI Engineer

“Expertise is accumulated, situated competence, and almost nobody is scaling it.”

AI Engineer

“Code is already a language native world, symbolic and structured, with tests standing in for rewards. The rest of digital work is millions of micro worlds, each with its own local physics, far too heterogeneous for one static model to compress.”

AI Engineer

“scale intelligence alone and you get what he calls the world's smartest novice, brilliant at whatever is put in front of it and accumulating nothing between problems”

AI Engineer

“if continual learning gets good enough past some threshold of raw capability, the thing worth scaling stops being the model”

AI Engineer