A basic challenge in education is helping knowledge transfer to new situations by seeing deeper structures -- pattern recognition, in other words. LLM models are still struggling mightily with this.

'But the idea that a machine or a chatbot can actually teach as a human can' represents “a profound misunderstanding of what AI is actually capable of,' says the person who led these efforts for more than a decade at IBM (using Watson). Really great reporting here.

A nice list that celebrates the wonder and depth of human cognition.

Point 1 is what Cognitive Resonance aims to do. Point 4 is why it exists.

'In fact, the National Council of Teachers of English recently announced its support for the idea of 'decenter[ing] book reading' in English language arts education.'

Wait, what?

Make no mistake, 'personalized learning' is being revived under the banner of AI. I share Abeba Birhane's concern about the many ways in which this can -- perhaps will -- go wrong. (And if you care or worry as well, follow her.)

If nothing else, I appreciate the implicit concession that *current* AI systems cannot reason, nor can they plan.

Resistance to the overhyping of AI in education is not futile open.substack.com/pub/buildcogni…

'Undersold' is a tough place to be right now, glad I'll be on the other side of this debate. If you're attending ASU+GSV next week, hope you'll join this conversation in person.

The researchers and authors of one of the best Substacks on AI have a forthcoming book. Note the title!

This is so bad.

Reddit upvoting, but for morality using AI.

2030 version: You don't need to buy any sweeping theories about [AI doomerism] to see that [an essay-generating app] is bad for getting teenagers [to learn how to think and write essays]...

One of the sharpest thinkers on the intersection of cognitive science and artificial intelligence is Sean Trott, and he's been kind enough to participate in an extended dialogue with me on what the future may hold. Here's part one of our conversation: buildcognitiveresonance.substack.com/p/do-we-need-a…

And yet, crossword puzzles exist.

The discussion here between Noah Giansiracusa and François Chollet regarding general intelligence in humans is illuminating.

Evaluating students using tests is hard. Evaluating AI using tests? Just as hard, maybe harder. Some thoughts on the challenge of using benchmarks to capture the complexity of intelligence, with an appearance by Grover from Sesame Street.
buildcognitiveresonance.substack.com/p/what-grover-…

What Grover and Good Will Hunting can tell us about the limits of artificial intelligence (my ode to research by Deb Raji @[email protected] on Mastodon Amandalynne Paullada Alex Hanna (اليكس حنٌا) & Emily Denton) open.substack.com/pub/buildcogni…

I'm reading through a government-issued report on AI in education and it is chock full of amazing graphics such as this one.

Once you learn a bit about how generative AI works, 'hallucinations' like this start to make sense. Colin Fraser | @colin-fraser.net on bsky has been a big contributor to my own understanding of these models.