Watch now (56 mins) | How can AI help with learning and education? And does it pose an extinction-level threat to the teaching and assessment models that currently dominate schools and universities?
I can't help thinking of a comment I read recently:
We thought people were stupid because they didn't have access to information.
Now we have the Internet.
We were wrong.
The major reason people don't learn is not that they don't have access. It is that they are not interested. They are interested in a credential, or often, knowledge that they actually know will be useful. That does not describe a lot of schoolwork.
I especially appreciate the discussion that AI is laying waste to the poor education system we have built of late. In particular, it exposes that what we have convinced ourselves is so important (e.g. memorization without obvious purpose, wrote arithmetic skills without understanding, non-creative writing) is actually trivial in the end.
Higher Ed is about to be shaken to it's foundation..
If we really value creativity, problem solving, statistics, and the like (I certainly do), how well can we teach that?? Or, better, is AI going to help?
I think the struggles described in the conversation about adopting AI in companies speaks to this challenge. Or maybe the simple failure of currwnteducation to do this. Or both! 😉
LLMs have been an absolute boom for my ability to very efficiently acquaint myself with unfamiliar disciplines. My day job requires me to consult with bona fide field experts and that's been an amazing proving ground for my self-guided tutoring. For example, as a criminal defense attorney, I'm supposed to know a lot about DNA analysis, and my knowledge gets tested against antagonistic expert witnesses.
And why not use LLMs to keep the take-home essay relevant? All you have to do is feed the student's paper into chatGPT and ask it to generate a 5-minute quiz that can discern that the student actually understands the material they wrote about. That's a grading exercise that shouldn't take much effort but it would nevertheless be very effective at separating the wheat from the chaff.
A minor comment on one point: "Another downside with in-person vivas and exams, which I hear particularly from high school teacher friends, is they’re just very labour-intensive. Compared to take-home essays, running exams regularly is classroom time where you’ve got to have a teacher in the room, where students are not learning."
In my experience, essays are the most labour-intensive form of assessment, because the teacher needs to read and evaluate all of them! Which is one reason why some teachers have already shifted from essays to cheaper assessments like MCQs and short open questions. This applies equally to whether essays are written at home or in class, as correction time is far greater than in-class time (and furthermore teachers don't get much credit for all this correction time, usually only their class time is counted). So I can't see the ditching of the take-home essay as creating a resource problem.
However, I entirely agree that this creates a huge problem for the development of writing skills.
The number of conversations almost exactly like this at many universities is good to see -- unfortunately none are happening at the administrative level. As an ex-dean I know where these conversations *should* be happening but aren't.
The danger is people not knowing themselves enough to use it wisely.
If education helped individuals develop genuine awareness, emotional maturity, and intellectual independence, AI wouldn’t be a threat, but a multiplier of human potential.
Thanks for the Anki tip! I’m a medical student so I use it a lot and never thought to have AI make me decks. In my experience medical school is one exception where AI is simultaneously extremely useful and it’s mostly not possible to use it to cheat
How does the use of AI to write an essay differ from the use of a paid human? Or is it only the removal of financial barriers to the cheating option that concerns you? And how long a face-to-face interview with the student do you think would be needed to identify a probable lack of authorship?
Yeah it costs nothing to produce an ai-generated essay and they can be much better and more tailored as well. Face to face interviews can help but there are so many barriers to organising them, including costs and the institutional challenges you run into if a student rejects an accusation.
Quite an eye-opening conversation. I'm a retired teacher (living in France) who used to teach philosophy (including logic) way back in the 90s and early 2000s (at the Univ. of Colorado @ Boulder among other places). I'm quite astounded by your impression that academics have taken little interest in AI. How can this be? It's all around us. I would have thought they would have been all over this. Personally I love AI, and I use GPT (and sometimes Gemini) essentially every day for my own personal educational goals. As my brain is aging, GPT has become my essential thinking companion and has helped me a great deal in relation to my own little hobby of modeling logic and probability puzzles with Bayesian networks. Were I to teach logic again, I would most definitely incorporate Bayesian Networks in my teaching toolkit. While GPT is aware of how Bayesian networks function, and how to build them, it's not fully able to completely and autonomously build them, but I have found GPT an absolutely invaluable helpmate towards completing my personal projects. There are things that I now try to tackle that I otherwise would not have in the absence of a reliable guide in the background because that would have been too complicated to do alone. So is one possible answer that teachers can be even more demanding about what students should be able to learn knowing that AI is always there when they are stuck? I don't know. I hope you periodically return to this issue and keep us updated to how this progresses.
Absolutely fantastic conversation - you cover the pros and the cons but I'm especially appreciative that you know what you're talking about when it comes to the technology. So many opinions about AI are from people who have not touched it or aren't aware of the speed of improvements. Great discussion.
Very informative and your use of AI is impressive. I still feel you've not answered what is in a way the most fundamental question posed by AI, which I'd put like this: In 2150, who will be the educated person and what will they know or be able to do that less educated persons will not be able to do. (For example, we might say that today's educated person will understand the strengths and weaknesses of Chat GPT, will have a grasp of the history of their own country, will have an appreciation of the masterworks produced by their society in literature, music, drama, and so on). I realize it may not be possible to answer the question at this point in time.
I can't help thinking of a comment I read recently:
We thought people were stupid because they didn't have access to information.
Now we have the Internet.
We were wrong.
The major reason people don't learn is not that they don't have access. It is that they are not interested. They are interested in a credential, or often, knowledge that they actually know will be useful. That does not describe a lot of schoolwork.
Excellent piece, gentlemen.
I especially appreciate the discussion that AI is laying waste to the poor education system we have built of late. In particular, it exposes that what we have convinced ourselves is so important (e.g. memorization without obvious purpose, wrote arithmetic skills without understanding, non-creative writing) is actually trivial in the end.
Higher Ed is about to be shaken to it's foundation..
But..
If we really value creativity, problem solving, statistics, and the like (I certainly do), how well can we teach that?? Or, better, is AI going to help?
I think the struggles described in the conversation about adopting AI in companies speaks to this challenge. Or maybe the simple failure of currwnteducation to do this. Or both! 😉
LLMs have been an absolute boom for my ability to very efficiently acquaint myself with unfamiliar disciplines. My day job requires me to consult with bona fide field experts and that's been an amazing proving ground for my self-guided tutoring. For example, as a criminal defense attorney, I'm supposed to know a lot about DNA analysis, and my knowledge gets tested against antagonistic expert witnesses.
And why not use LLMs to keep the take-home essay relevant? All you have to do is feed the student's paper into chatGPT and ask it to generate a 5-minute quiz that can discern that the student actually understands the material they wrote about. That's a grading exercise that shouldn't take much effort but it would nevertheless be very effective at separating the wheat from the chaff.
Great conversation!
A minor comment on one point: "Another downside with in-person vivas and exams, which I hear particularly from high school teacher friends, is they’re just very labour-intensive. Compared to take-home essays, running exams regularly is classroom time where you’ve got to have a teacher in the room, where students are not learning."
In my experience, essays are the most labour-intensive form of assessment, because the teacher needs to read and evaluate all of them! Which is one reason why some teachers have already shifted from essays to cheaper assessments like MCQs and short open questions. This applies equally to whether essays are written at home or in class, as correction time is far greater than in-class time (and furthermore teachers don't get much credit for all this correction time, usually only their class time is counted). So I can't see the ditching of the take-home essay as creating a resource problem.
However, I entirely agree that this creates a huge problem for the development of writing skills.
The number of conversations almost exactly like this at many universities is good to see -- unfortunately none are happening at the administrative level. As an ex-dean I know where these conversations *should* be happening but aren't.
The danger is people not knowing themselves enough to use it wisely.
If education helped individuals develop genuine awareness, emotional maturity, and intellectual independence, AI wouldn’t be a threat, but a multiplier of human potential.
Thanks for the Anki tip! I’m a medical student so I use it a lot and never thought to have AI make me decks. In my experience medical school is one exception where AI is simultaneously extremely useful and it’s mostly not possible to use it to cheat
How does the use of AI to write an essay differ from the use of a paid human? Or is it only the removal of financial barriers to the cheating option that concerns you? And how long a face-to-face interview with the student do you think would be needed to identify a probable lack of authorship?
Yeah it costs nothing to produce an ai-generated essay and they can be much better and more tailored as well. Face to face interviews can help but there are so many barriers to organising them, including costs and the institutional challenges you run into if a student rejects an accusation.
There's no need to make an actual accusation. Just make the grade for the essay dependent on the student's ability to defend it in person.
Quite an eye-opening conversation. I'm a retired teacher (living in France) who used to teach philosophy (including logic) way back in the 90s and early 2000s (at the Univ. of Colorado @ Boulder among other places). I'm quite astounded by your impression that academics have taken little interest in AI. How can this be? It's all around us. I would have thought they would have been all over this. Personally I love AI, and I use GPT (and sometimes Gemini) essentially every day for my own personal educational goals. As my brain is aging, GPT has become my essential thinking companion and has helped me a great deal in relation to my own little hobby of modeling logic and probability puzzles with Bayesian networks. Were I to teach logic again, I would most definitely incorporate Bayesian Networks in my teaching toolkit. While GPT is aware of how Bayesian networks function, and how to build them, it's not fully able to completely and autonomously build them, but I have found GPT an absolutely invaluable helpmate towards completing my personal projects. There are things that I now try to tackle that I otherwise would not have in the absence of a reliable guide in the background because that would have been too complicated to do alone. So is one possible answer that teachers can be even more demanding about what students should be able to learn knowing that AI is always there when they are stuck? I don't know. I hope you periodically return to this issue and keep us updated to how this progresses.
Absolutely fantastic conversation - you cover the pros and the cons but I'm especially appreciative that you know what you're talking about when it comes to the technology. So many opinions about AI are from people who have not touched it or aren't aware of the speed of improvements. Great discussion.
Very informative and your use of AI is impressive. I still feel you've not answered what is in a way the most fundamental question posed by AI, which I'd put like this: In 2150, who will be the educated person and what will they know or be able to do that less educated persons will not be able to do. (For example, we might say that today's educated person will understand the strengths and weaknesses of Chat GPT, will have a grasp of the history of their own country, will have an appreciation of the masterworks produced by their society in literature, music, drama, and so on). I realize it may not be possible to answer the question at this point in time.
Excellent conversation, especially about what is happening at the high school level. Higher ed is simply not prepared.
Thanks 🙏. Yes, it's worrying!