Stephen Downes

I think what boths me most about both proponents and critics of AI in education is that none of them can imagine any other model than an instructor teaching a student. It's like both proponents and critics are frozen by this image of a computer screen - an oblong box - as though no other form of interaction could ever exist. It's Taylorism, not technology. I mean, here's Ben Williamson: "despite the rhetoric of transformation, all these AI tutors really seem to promise is a one-to-one transactional model of learning where the student interacts with a device. It's an approach that might work OK in the staged setting of a promo video recording studio, but is likely to run up hard with the reality of busy classrooms." Well, yeah, if a "busy classroom" is your ideal of learning, AI tutors might fall short. But that's like saying airplanes won't fit through railway tunnels.

This article questions the thinking behind "contingency plans for universities in the event that they fail which would protect researchers and their research," stipulating a set of conditions (eg., "there is sufficient funding", "the researcher wishes to move", etc.) needed to make this happen. But while not endorsing the contingency plans I would suggest that the heading really should be "You can't save university research without saving universities". I mean, I work in a non-university research institute, and I know of many others. Take, for example, the comment that "a building I know only too well cost ~£55M to construct, uses ~£1.5M+ in electricity a year and produces some very important technological advances." Nothing stops it from continuing on its own. There's no reason a full university infrastructure needs to surround that building, and in the case of many stand-alone research institutes world-wide, it doesn't.

Daniel Christian summarizes three articles looking at the recently released AI models. Though I think the lead article on LearnLM from AI Supremacy isn't useful (it feels like an AI-assembled jumble of acontextual platitudes) the second makes an important point: "This tool (ChatGPT4o) is ideally positioned not only to instruct but also to sell products, shape minds, and manipulate real world events." The third is also interesting, but in a different way, as it makes the claim that free products are "market distortions". For example, "the entire LMS market of the 2000s sagged under the weight of Google Classroom's $0 price tag." But - importantly - there's no such thing as a non-distorted market. There's no 'natural state' of the market - that's just fantasy. The very concept of the market entails people trying to 'distort' the market by manipulating either supply or demand.

I really disagree with the main argument in this post, though I imagine many readers (especially those that are educators) would agree: "Whether it's a project-based learning unit or a class discussion, it is the teacher, as the artist and the problem-solver and curator, who sparks innovation." While this reads as though John Spencer thinks that the AI is incapable, it is also saying the same thing about students. This perspective places most of the agency in the hands of the teacher and almost non in the hands of the student. We did it that way because teachers were scarce. But when we can have individual teachers for each person, we need not depend on the teacher to set a direction. And I think we'll find that individual students - especially with AI support - will be able to cope.

The story: "The Verge and 404 Media are building out new functions that would allow them to distribute posts on their sites and on federated platforms – like Threads, Mastodon and Bluesky – at the same time. Replies to those posts on those platforms become comments on their sites." Sounds great, but I've been blocked by a a paywall on 404 in the past, and this doesn't really work with an open fediverse. So there's going to need to be some thinking about how all this works (I'm also considering how a similar plan for OLDaily would work, though the number of comments on OLDaily posts is pretty minimal - it's more of a design exercise for me).

Overview of the ways AI-supported learning will be able to draw on context. "Where AI has been lacking is in knowing about your immediate context and intent. This nut is now being cracked, as AI's multimodal abilities have now hit the market. By this I mean its ability to 'hear' things, 'see' through your camera, identify things from video or know what's on your screen."