Stephen Downes

This is a set of interviews with five experts about whether AI can ever match human intelligence. The interviews are relatively short and easy to read, and the content is well-informed. The experts mostly default to the view that 'machine intelligence and human intelligence are fundamentally different' and while they offer explanations of how they differ (eg., as Emily Bender says, in attaching meaning to words) we don't really get any insights on what makes them difference. I suspect that this is because they don't know (eg., how do humans actually attach meaning to words?). My own thought is that human and machine intelligence are fundamentally the same, with limited differences based on the nature of their physical composition.

Here we have what I think is a great example of a course where it doesn't matter whether you're using AI. It's based on building your own clocks with microelectronics, and as such, hands-on in a way the AI cannot substitute (not, at least, until we get very capable robots). An advantage of this course it that it depends on stock low-cost components easily shipped to your home from the internet. "For under $15 in parts — a Raspberry Pi Pico 2 W, a 2.1" round color display, three buttons, and a breadboard — students build a real, ultra-smart clock from scratch. No soldering required."  Via Dan McCreary.

Tom Watson argues in favour of seeking out productive friction. "The often quoted Thinking and Fast and Slow talks of two types of thinking System 1: fast, intuitive, automatic thinking and System 2: slower, effortful, deliberate thinking. The removal of friction from much of our thinking work means we need to be very deliberate in leaning into System 2 as a practice." My own preference is to use the word 'resistance' rather than 'friction', and in slogan form, it's "Do hard things."

According to this article, "the race is shifting from 'Who has the smartest chatbot?' to 'Who can build an AI that helps a person learn without doing the learning for them?'" It's a good question, because it's not exactly clear what the answer would look like (and indeed, it might be different for each student). "The problem is timing. If a beginner uses AI to bypass the exact mental operation they are trying to learn, they may become efficient before they become competent." This would have been a stronger article had it stuck to this question, but the AI-assisted text wanders as it goes on, talking more about the role of the teacher and that the AI can, and cannot, replace. "The winning model is unlikely to be 'AI teaches, humans watch.' It is more likely to be a deliberately designed triangle: the student does the learning, the AI supplies adaptive support, and the teacher decides what the learning is for."

According to this article, "The Compass provides a method for thinking about governance and implementation of technologies holistically to help education systems navigate technological change while keeping human rights, democracy and the rule of law at the centre." The 'compass' has two major axes: "the governance of people and the governance of processes"; and "education with AI and education about AI." There are four layers: exploration, integration, implementation, and transformation. Around the outside are eight interconnected systems. The intent is fir the compass to be a guide, not a destination, and it's a nice metaphor, but I'm not sure it advances discourse in any particular way.

This article takes a problem of coordination, translates it into a problem of ownership, and recommends a centralized solution in the form of a Chief Online Learning Officer (COLO). It sounds great, but it seems to me that it just pushes the problem deeper (and more intractably) into the different departments. Jeremiah Grabowski references decisions about online learning being made in academic affairs, instructional design, admissions, and marketing, among others. Is the COLO now going to tell marketing how to market? Admissions who to admit? This simply creates contradictions in each of those departments, where they no longer have the mandate to manage what they have the putative mandate to manage.