From Skeptic To Believer
Dylan recalls moment his AI skepticism cracked
Jun 24, 2026 · 1 hr 28 min · 12 segments
In this AI in Education episode, Craig Barton welcomes back Dylan Wiliam to discuss what generative and extractive AI mean for assessment, teaching and the purpose of schooling. Wiliam argues the…
Dylan WiliamGuest
Craig BartonHostNo entities detected.

So before we get into the nitty-gritty of how AI can play out in assessment and teaching and learning, I'm interested in your personal kind of AI use cases.

And if we go back to when ChatGPT launched, were you super excited? Were you skeptical? And, and how would AI play out in your typical day-to-day, Dylan?

I was skeptical, uh, partly because there'd been so many false dawns previously.

You know, people have always said this time is different, but I really think it is this d- different this time.

We are seeing extraordinary s- uh, leaps, and I've stopped saying AI will never be able to do X because I've said that so many times and been wrong.

I very quickly discovered, uh, a guy named Ethan Mollick, a professor at The Wharton School at the, at the University of Pennsylvania in Philadelphia.

And he's done a lot of work to make it really easy to p- for people to, to use AI in productive ways.

S- to begin with, very trivial things, like formatting references that came in a non-standard form.

So I'd drop a, a research paper, and I just said, you know, "Give me the, give me the reference in a .ris format.

Um, give me a, an abstract of the paper that, that's different from the paper that's, p- the abstract that the authors wrote." And so I used it for small things like that.

Um, I then started using it for interesting things like, um, uh, looking at student work.

So I- looked at some of this project work I've been doing back in the 1980s on students' logical thinking.

You've got, uh, uh, four squares together, and you give students four colors, and you ask them, "How many different ways are there?" Not counting rotations and reflections are different.

How many ways are there of coloring this shape with four colors? And what I was quite impressed by, this is GPT-4, I think.

It was able to tell me that the two classes had adopted, adopted different approaches.

So before we get into the nitty-gritty of how AI can play out in assessment and teaching and learning, I'm interested in your personal kind of AI use cases.

And if we go back to when ChatGPT launched, were you super excited? Were you skeptical? And, and how would AI play out in your typical day-to-day, Dylan?

I was skeptical, uh, partly because there'd been so many false dawns previously.

You know, people have always said this time is different, but I really think it is this d- different this time.

We are seeing extraordinary s- uh, leaps, and I've stopped saying AI will never be able to do X because I've said that so many times and been wrong.

I very quickly discovered, uh, a guy named Ethan Mollick, a professor at The Wharton School at the, at the University of Pennsylvania in Philadelphia.

And he's done a lot of work to make it really easy to p- for people to, to use AI in productive ways.

S- to begin with, very trivial things, like formatting references that came in a non-standard form.

So I'd drop a, a research paper, and I just said, you know, "Give me the, give me the reference in a .ris format.

Um, give me a, an abstract of the paper that, that's different from the paper that's, p- the abstract that the authors wrote." And so I used it for small things like that.

Um, I then started using it for interesting things like, um, uh, looking at student work.

So I- looked at some of this project work I've been doing back in the 1980s on students' logical thinking.

You've got, uh, uh, four squares together, and you give students four colors, and you ask them, "How many different ways are there?" Not counting rotations and reflections are different.

How many ways are there of coloring this shape with four colors? And what I was quite impressed by, this is GPT-4, I think.

It was able to tell me that the two classes had adopted, adopted different approaches.
Every episode on Radar is fully transcribed, speaker-labeled, and rich with metadata. Here is a taste of this one. Try Radar for free to see the rest.
3 of 7
From Skeptic To Believer
Dylan recalls moment his AI skepticism cracked
AI Beats Doctors’ Empathy
AI beats humans at creativity and empathy
AI Coach In The Classroom
AI coach offers teachers personalized physics feedback
+4 more clips · 6 min 6 sec of audio in all
7 of 21
The rest of this transcript — segmented and speaker-labeled, so you land on the exact moment something was said
All 7 clips — the highlight moments, each cut as its own audio, with a title and a speaker
All 12 segments — the transcript broken into labeled sections, every ad read marked
All 21 topics — jump to every other episode discussing the same subject
Every related episode — other shows Radar links to this one
No account is needed to search Radar.