Stephen Downes

If you used AI to write some software, who owns the result? The laws are still being clarified but are settling along the lines of a 2021 recommendation from the Canadian agency Innovation, Science and Economic Development Canada (ISED): "Ownership belongs to the person who arranged for the work to be created. Ownership and copyright are only applicable to works produced by humans, and thus, the resultant code would not be eligible for copyright protection. A new 'authorless' set of rights should be created for AI-generated works." Of course it's not quite that simple, because there's also the question of who is liable for 'authorless' code. And also, most AI-generated code is hybrid: the AI produces a concept, but it is then worked over by the human to tailor the result.

Simon Willison documents his work with Claude AI as he builds a like-blogging took in Django, a Python framework. Here's what he wanted: "Write JavaScript (no React) that polls /updates/1/ using fetch() and takes the HTML from that and injects it into innerHTML in a div with id='updates'" (take not of the expertise required to tell Claude exactly what you want). What gets me is he built this trool while he was live blogging a conference. During the lunch break, he made some improvements. He also switched over to ChatGPT 4o as he continued his built. He built it back in October and used it again this week.

"When we look back at 2024," writes Carlo Iacono, "I believe it will be remembered as the year education started to embrace reality over illusion." Despite assertions that we've hit the 'AI wall' the difference between what AI could di this time last year and what it can do to day is like night and day. "If your last serious engagement with AI was ChatGPT 3.5 eighteen months ago, or even an early version of ChatGPT-4 or lets be honest any version of Microsoft Copilot (apart from the dearly departed Sydney), you're judging a cathedral by its foundations." I agree. There's still a lot of room for AI to improve, and it will very likely do so.

What would we have gotten had music publishers owned the radio stations back in the day? Probably something a lot like Spotify. This article explores how the publishers outsized influence over the company has pushed income for artists lower and lower to the point where some are getting a small payment for an anonymous track, all rights belonging to the publisher. There's a warning here for the content industry writ large. (Harper's allows you one free view, here's a Wayback version just in case). More.

There has been a 'blog-or-die' meme circulating around, which is good for me, because it gives me more things to pass on in my newsletter. Not that Alan Levine needed a push! Anyhow, here he is talking about podcasting. He links to "the massive and open API enabled Podcast Index where you can search and find almost any podcast" and visit's Ed's podcast list, which I linked to a few days ago, and offers his own Cogdog's Podcast List (also good for me because it gave me Ed's last name). Thus inspired, I exported the OPML from my own podcast player to bring you Stephen's Podcast List, a set of favourites curated over decades of online listening.

This starts out as a good post but gradually decreases in quality with each successive paragraph, finally resting on the well-worn Socratic critique of writing, "forgetfulness in the minds of those who learn to use it." The main (and interesting) point addresses the relevance of the question in the title in the first place: "Such questions matter (not because of any answers they may or may not provide), but because their framing is indicative of a techno-determinist mindset, which is itself inseparable from the marketing of edtech products." Why? "The only meaningful question concerns how the technology is used in specific settings with specific groups of learners," which is something, says Philip Kerr, that a meta-analysis can never address (why not? we'll never know...). Alas, instead of pursuing what might have been a fruitful discussion, we are taken on a tour of poor and increasingly threadbare allusions (a line of coke, mental effort, cognitive load, The Shallows, profit-driven corporations, Socrates).