In the introduction to this session Rob Reich highlights the intensely decentralised nature of American higher education and what it means for the adoption of technological innovations across the sector. This is extremely different to the UK sector which is (weirdly) hyper-centralised but also (for now) differentiated. Exactly what this means for the diffusion then institutionalisation
This takes me back to categories from Margaret Archer in her Social Origins of Education Systems where she distinguishes between centralized/decentralized (structural characteristics) and systematization/differentiation (cultural characteristics) of educational systems.
Percy Liang’s contribution is extremely clear about the need for foundation models (which I think a broader category than LLM, in the sense that all LLMs are foundation models but not vice versa) to be pedagogically scaffolded because their imperative is speed and accuracy rather than supporting learning.
There’s a lot of discussion here about the analogy between generative AI and a calculator. Rob Reich’s point is that a calculator isn’t an agent in the classroom in a way that AI has the potential to be. In contrast the panel suggest that the currency on ‘getting text out’ will evaporate in the same way as the currency associated with doing arithmetic; the locus of aspiration will shift from the brute production of words to the thinking which underpins it and iterative editing which follows from it.