Stephen Downes

This tool, which the website source reveals was created by Perplexity computer, compares 20 AI literacy frameworks across 9 domains. The domains are: technical understanding, practical application, critical evaluation, ethical reasoning, societal and systemic awareness, human agency and identity, governance and participation, cognitive and metacognitive processes, and sociocultural and critical orientation. My question is: what makes these good dimensions of AI literacy, over and above the fact that they may be extracted from the 20 literacy frameworks viewed? Minimally, a typology should be comprehensive and non-overlapping. But more to the point, what exactly is a 'literacy'? My own view is that a 'literacy' is a type of pattern recognition. What alternative characterization would a presentation like this offer? It seems circular - a 'literacy' is what people say is a literacy.

Colin Beer is usually sharper than this, so while I agree that knowledge and skills (as he defines them) are not enough, I think we need some clarity regarding what he calls 'dispositions' (it's not that he's wrong so much as he's fuzzy). He writes, "Dispositions represent the values, tendencies, and attitudes, such as motivation, mindset, professional identity and agency, that dictate how a professional actually navigates the "swampy lowlands" of practice. In simple terms, dispositions are the habits of mind and heart that shape how we show up when work gets hard." Dispositions are best described as tendencies, which may result from habits, or which may be subconscious tics. They should be contrasted with attitudes, which are states of mind regarding such things as values and truth. Expertise (in, say, the Dreyfus sense) is a matter of disposition, while professionalism is a matter of attitude. It's certainly arguable that an education should (help) shape both, but they are very distinct things, and are approached very differently.

I spent some time thinking about this short article that explores "the difference between good relationships and good transactions" where a 'good relationship' is "unique, organic, and empathetic, helping us understand when to invest in building relationships versus when a transaction suffices." What made me think wasn't the distinction itself, which seems straightforward, but the terminology used. The way 'relationship' is defined blends elements of different constructs - we have 'unique' and 'sustained', which to me describes a 'connection', but in addition there is the presumption that relationships are embodied, as evidenced by 'organic' and 'empathetic'. The connection describes the relationship itself, while the embodied element describes the thing that is related. The transaction side, meanwhile, describes the exchange that happens between two entities, as opposed to the connections between them. The world view of this article doesn't grant (or doesn't require?) embodiment for transactions to occur. I would ask whether the author intended to distinguish between embodied and non?-embodied entities here, or whether it's just phrasing.

The current issue of The Batch introduces readers to Context Hub, Andrew Ng's new tool to provide API documentation to your coding tools. The purpose of this is to make the tool aware of new tools and updates to existing tools, so they're not depending on out-of-date models. "Chub is built to enable agents to improve over time. For example, if an agent finds that the documentation for a tool is incomplete but discovers a workaround, it can save a note so as not to have to rediscover it from scratch next time." It's available on GitHub and installed using the node package manager (npm). It's the lead article in this issue; you can also read other AI news from the week.

This is a good article, I'll grant it that. I resist its main thesis; ultimately the argument is not successful. But it's worth state here. The thesis - as suggested by the title - is that life is inherently different from non-life. "Organisms are more than just machines, and minds are more than just computers." The main argument, which Adam Frank draws from Hans Jonas, is that "living systems are not stable collections of atoms like a rock. Instead, they are stable patterns that persist through time... a specific kind of organization through which matter and energy pass." And because life is a type of organization, and not reducible to matter and energy, it has special needs, for example, "interiority and individuality." Also, "every organism must actively maintain itself against the continuous threat of its own dissolution" and "life always has purpose." There is additionally the argument from Robert Rosen that "metabolic systems could be viewed as a special kind of organization where networks of processes close back on themselves" and hence "not Turing computable."

This is a clear and well-written account of what an application programming interface (API) is. We read about APIs all the time, from xAPI for learning records to MCP as an API for artificial intelligence. This article describes in an accessible way what we mean by API, exactly. "When engineers build modules of code to do specific things, they clearly define what inputs those modules take and what outputs they produce: that's all an API really is."