Stephen Downes

It's as though people believe there's a learning 'off switch' (though it only exists in other people). Like this: "A junior who made a mistake is one step closer to being a senior; a junior who let an LLM make a mistake (and had the LLM fix it for them) has probably learned nothing." What would justify this conclusion, authored by John Collinsworth (Nielsen doesn't link to it, though he quotes it extensively - boo, hiss)? The junior will still learn, but will learn a different thing. Nobody 'learns nothing' - human brains don't shut off like light bulbs. What's really happening here is that we're mking a value judgement, specifically, that the lesson learned from doing it by hand is more important than the lesson learned by doing it with AI. This 

The first think I thought of when I read this post from Julian Stodd was simulated annealing and Boltzmann machines. Why? Because Stodd describes crystals as structures that "represent the lowest energy configuration," which is what Boltzmann machines try to do with neural nets. Stodd compares crystals to organizations, and here the metaphor needs rescuing a bit. He writes, organizations "build structure for greatest efficiency, value, predictability, replicability, and of course, ease of control, and potential for oversight and measurement," but that they are "systems trapped at one energy level, dreaming of the next." Sure, they can change state - they can melt into fluid and merge with something else, or they can sublimate into their individual atomic components (a.k.a. people). But they can change energy level, but (just as with matter and neural nets) they need to go through an annealing process, a 'hardening through fire', as it were. 'Move fast and brek things' is a clumsy attempt at just such a process; in neural networks they just increase and decrease the bias (ie., sensitivity, not prejudice (that's a different meaning of 'bias')) of individual entities.

What's great about this article is that it really clearly defines the distinction between Mastodon and Bluesky, arguing that the question "where are the instances?" is a Mastodon question, and doesn't really apply to Bluesky. But I think it's misleading in three major ways. First, there are instances in Bluesky, but there are different types of them: 'atproto hosting' and 'apps'. The fact that there are so few 'atproto hosting' instance is an issue. Second, it under-describes the ATmosphere architecture by leaving out the 'personal data store' (PDS) entirely. And third, it argues that atproto is just like RSS and Google Reader. But RSS is a simple text file any person can create in a text editor (so, for that matter, is a PDS, sort of), while a 'atproto hosting' instance is a very large and complex piece of software that consumes massive resources. So too are the apps, which for some reason, must handle "the whole Atmosphere". And the whole argument raises the question: why can't we simply have PDS readers as apps? What is the whole 'atproto hosting' infrastructure buy us, except control over our identity?

This is a longish discussion of McLuhan's idea of a tool as an extension of the self, which then of course can be applied to media. Today, it can be applied to AI. So when we say 'AI is a tool' we're not limiting our imagination - "the ceiling of what tool literacy can imagine" - we are instead inviting ourselves to explore how a human and an AI can be (dare I say it?) a system. Now I'm not sure Levine meant his discussion exactly that way, but this is what follows from the text. And I question both McLuhan and that reading. I used to consider the question of whether a social network is an extension of the neural network (which is essentially the Siemens interpretation of connectivism). But I see them as two separate networks, with a process of communication (essentially, emergence and recognition) between them. The McLuhan argument is a nice metaphor. But it should not be taken literally.

The headline here is that "a year after 250 CEOs demanded mandatory AI education, industry leaders are zeroing in on the durable 'human' skills they can't hire for." That sounds good, but as always, I would warn against depending on industry leaders to define what should be taught. For one thing, it's often wrong. But more importantly, their advice is intended to benefit them, not the students. Consider this: "what we need is to figure out how to teach the human skills – how to teach future-proof skills that set an employee up for success no matter what domain they find themselves in." Why would this be important. Well, it could be because technology is changing so rapidly. But from where I sit, it's just as likely that employers want to hire human cogs they can quickly 'retrain' and slot into positions the employees never expected to be doing when they were hired - a lot like the way Facebook transferred software engineers into data-labeling peons.

I'm not sure 'disjuncture' is the best word for the contradictions being described by Glenda Morgan in this article on the differences between what teachers think they are delivering and what students think they are receiving. But the point is still valid. For example, "When asked whether they were incorporating real-world projects into their courses, between 58% and 73% of faculty said they were, depending on whether workforce readiness was a high priority for them. Yet only 26% of students reported completing a real-world project in a course." Considering the difference in employment between those who did, and did not, receive workplace experience, you can see ho important this is.