It is not surprising that there is growing scepticism and even open opposition to AI in education. Even if the latest release of Large Language Models have shown a marked improvement Generative AI remains problematic on a number of levels including so called hallucinations, ongoing biases, limited provision in many languages and so on. Perhaps most problematic is the control asserted by a very limited number of large manly American companies all of whom have tried to hype. The value of their products for education.

All of which is true but is just a continuation of the trend to over hype the potential transformative impact of technology in education which goes back many years. And this seems linked to our preoccupation with classrooms at the centre of learning. Phillipa Hardman, a researcher and developer in corporate learning, claims we have built an Industry around the Margins. She points out that most learning time is not formal. She draws attention to the classic 70:20:10 work by Lombardo and Eichinger (1996) that found that “people attributed roughly 10% of their development to formal learning, and 90% to informal and experiential learning — work itself, stretch assignments, peers and mentoring.“ Looking at learning for work she says “transfer-of-training research suggest that only around 10–20% of formal training leads to sustained behaviour change on the job (Baldwin & Ford, 1988; Henao-Calad, 2024). And she questions whether participants who have passed courses can really do what the assessment results say.

But this is the market that the big tech AI companies are going after, not because they think that this is where Generative AI can best support real learning but because this where their marketing people think the big money is. And with the bubble around AI looking increasingly vulnerable and business models increasingly unrealistic education is an attractive vista. Yet the increasing volume of research into the use of Gen AI within the existing model of education questions it pedagogic value for learning especially at a cognitive level. I wonder if AI might be better suited, not for competing with teachers in the classroom but supporting other form of learning. Indeed, one of the few areas of learning that AI seems to be doing well is in developing simulations. I like the approach by Andrek Karpathy to using a Large Language Model to incrementally build and maintain a persistent wiki - a structured, interlinked collection of markdown files that sits between you and the raw sources. “When you add a new source”, he says, “the LLM doesn't just index it for later retrieval. It reads it, extracts the key information, and integrates it into the existing wiki - updating entity pages, revising topic summaries, noting where new data contradicts old claims, strengthening or challenging the evolving synthesis. The knowledge is compiled once and then kept current, not re-derived on every query.” Using LLMs to develop Personal Learning Environments as much potential. But note that here the agency remains with the learner, not with the machine as it does with so called personalised learning. And this is just one idea. But if we could break the hold of the big companies and resource small language models and Open Source Software we can return to a focus on supporting learning, rather than privatising education.

About the Image

 "Server Pool" subverts the swimming pool of the Villa Empain, a Brussels contemporary art venue. The pool is drained of its water and filled with computer servers. The image makes visible the massive water consumption required to cool data centers, but also questions the allocation of resources: when enormous budgets shift toward digital infrastructure, what remains for culture? / Paper collage digitally recomposed. Created from Brussels heritage materials during a workshop organized by FARI – AI for the Common Good Institute Brussels.