Stephen Downes

This article examines a report from the Virginia Community College System (105 page PDF) and punctuates it with some relevant questions for college systems generally. The summary is pretty accurate and picks out exactly the diagrams I would have, including especially the huge drop in in-person only classes (and corresponding jump in online and hybrid course delivery) over the last ten years. Two of the three questions address this: "How do we deliver effective student supports as enrollment moves online? As demand shifts away from on-campus learning, do physical campuses need to get smaller?"

For those of you who don't remember Clippy, let me assure you that Clippy was... annoying. At first glance it appears that Mico will be no less so. "Mico is now being turned on by default in Copilot's voice mode, where you'll also have the option to turn the bouncing orb off." Ack. More, there's a Learn Live version of Mico that is "a Socratic tutor that 'guides you through concepts instead of just giving answers.'" Trust me, when I ask an AI something, I want an answer, not a dialogue about how my previous intuitions about the subject were wrong. And the Learn Live version of Mico distinguished by - get this - its glasses. Because of course glasses = teacher nerd, right? Related: Copilot rolls out group chats.

Back in the day, we used to have external hard drives. Now we're entering the world of external AI drives. It's made possible through devices like this AI workstation and connecting software such as the ZGX Toolkit that "streamlines AI workflows by pairing powerful local compute with open-source tools, built-in discovery, and easy export". Relevant systems in this category include HP's ZGX (more), Nvidia's Project DIGITS, Acer's Veriton GN100 AI Mini Workstation, and Lenovo's Mini AI Workstation. Prices are in roughly the $US 3000 range, which if history teaches us anything, will be in the $300 range for the same power in just a few years. You'll be able to put a few on a network and access them with your Raspberry Pi computers.

When we teach people about what words mean, what are we teaching them? Is the meaning of a word based on what the word refers to in the world? Is it based on some sort of convention about what such words mean? Or is the meaning of a word based on something else entirely? This paper looks at two philosophers who agree on an 'externalist' conception of meaning, that is, that meaning is based on what the word refers to in the world. But is there, as Saul Kripke would argue, a necessary connection between meaning and the object being referenced? Hilary Putnam offers a 'twin earth' argument where, say, we use the word 'water' in the same way, but where the actual substance of water isn't H2O, it's XYZ. Meaning could still depend on what we refer to, but it's not necessary that we always refer to the exact same thing. This is a fun paper describing these considerations and it offers the sort of argument that makes me think meaning isn't externally based at all. 32 page MS-Word. Image: Philosophy Now.

So I think we're about to be deluged by a ton of these AI-based learning platforms. They'll be designed along the same lines as the AI-based news services and even the new AI-based browsers - that is to say, they'll insert a layer of AI in between the user and the actual source. Tome Woodward is experimenting with that kind of thing here, as he uses Notebook LM as (shall we say) an 'interpreter' for an OpenStax textbook (is 'interpreter' the right word to use here? I haven't really seen any terminology that describes this, and 'interpreter' works in both an ordinary and technical context). There are many questions - does the AI add value, does it interpret the base text accurately, does it insert its own spin or marketing, does it enclose the public commons of knowledge and information? Woodward doesn't offer answer to these questions, but he's beginning the sort of exploration we need to undertake in order to get a sense of the whole domain.

OpenAI has released a new ChatGPT browser called Atlas. Not surprisingly, it has generated a lot of commentary, the most trenchent of which is this post from Anil Dash. It "actively fights against the web," he argues, and it's hard not to agree. "Atlas substitutes its own AI-generated content for the web, but it looks like it's showing you the web," he notes. And that's a problem. If I search for 'Tylor Swift', I want to see Taylor Swift's own web page, not an AI generated article about Taylor Swift. "The clear intent was that I was meant to stay within the AI-generated results, trapped in that walled garden." But it's not just that. Why are we typing queries into our browser at all? The great thing about the web is that we can click a link and be taken somewhere. "It's very clear," writes Dash, "that a lot of the new AI era is about dismantling the web's original design."