Is AI about to make education worse?

0:00the desire to learn things is essential to us as human beings and we need to make sure that education, the way formal education is structured, actually encourages that basic desire in human beings and doesn't discourage it by introducing anxiety. Today I want to talk about AI and whether AI can save education or at least make education more effective, more powerful. And obviously, AI is here to stay in different fields. People are looking at how they can make their field medicine or office management or whatever it might be more effective using AI. The problem is those people who are in education do they recognize that their model is ineffective. And if they apply AI to a model that doesn't work very well, it's not going to make that model much more effective. And that's basically the subject of my video today. So I was listening to an audiobook and reading on my Kindle. An audiobook called The Most Beautiful History of Intelligence,

1:01La Plumelle l'histoire de l'intelligence, by Stanisasse Leanne, Jan Lecun, and I think it was Jacques Girardon. Stanisasse Leanne is a person who writes about the brain and how the brain learns. Janic Lecun was the head of AI at Metta and is a professor and a very well-known expert and has many awards for his work on AI. And Jacques Girardon, as I understand it, is a journalist in the scientific field. So as a book, it's it popularizes the subject of how we learn, AI, and so on. Even a book that is popularized and therefore perhaps a little superficial, I don't fully understand it. But it gets me going in the subject. And I may later read something that's easier or more difficult on the same subject. But I start to gain a bit of a hold on the subject. And that form of learning has always been my approach to language learning. So I don't try to nail anything down. I don't use the sort of ladder approach, which is more of the

2:02traditional approach in education. You go to step one, you learn certain basic concepts, then arm with those basic concepts, which you are tested on. And if you pass, you move to the next level and the next level. And even at Duolingo, they force you to pass, as I remember, my brief exposure to Duolingo, you have to go to this first stage before you can go to the next stage. Whereas I have always had the approach that I would rather wander around exploring, sometimes things that are more difficult, or else going back to things that are easier. And again, I went to AI perplexity and asked about my experience of language learning and whether this has application in education more generally. And he said it does. And that the brain prefers to learn things, go back to something that the brain learned before, go over the ground again, and it in the end, it learns better in this way. So my feeling is with, as was the case with online learning, I saw online language instruction get going on YouTube and also, and yet a lot of what

3:06was offered was essentially the same old way of teaching languages just online. Whereas with text and audio and MP3 files, there was an opportunity to do something different based on how the brain learns, which is through massive input. So the concern here is, is AI simply going to reinforce the established sort of ladder approach to education, or is it going to enable this other approach where we kind of wander around this mountain on the mountain trail and we come across something which we notice and we kind of tuck that away for later. And then we come back again in a different context and we see the same thing. And through dealing with content that's more difficult or less difficult and encountering some of the same either facts or bits of information or structures in the language, we gradually learn them. And we don't need to demonstrate at each point that we have mastered anything. So that's more of a circular approach to learning as opposed to the ladder approach. But I asked AI, again, perplexity, I think, or Claude, what are teachers?

4:11How are educators planning to use AI? And as I kind of suspected, at least according to perplexity, the education establishment is hoping to use AI to develop better quizzes to improve their curriculum, but again, all based on the ladder vertical approach to education. I saw no evidence that they were planning to use AI in order to develop, for example, in math and social science, not dimension language learning, different varieties, different formats of the same basic content so that learners can experiment with, let's say, a podcast or a video or a simpler version or a more difficult version of the same basic material that has to be learned. And if they were to explore in this way, they would come back to the same material again and have an easier time of understanding it because they will then already have broken ground. They will already have some

5:13concept of what it's about. And that, rather than trying to sock something in, nail something down, master something, the assumption is I won't master anything. I'll just keep exploring. And at some point, more and more of the sediments of information will settle in my brain. Now, all this sounds fine. And I've mentioned this before in my videos, my approach to learning things. And of course, the question I always get is, so what about testing? We need to test parents want tests, educators want tests. How much has the student learned? Well, at least when it comes to language learning, I don't think that really matters because, as I've said, in the Canadian school system, kids learning French in the English language school system are tested all the time and at the end of their 10 years of learning French, they can't be French. So the testing really doesn't prove much. But at some level, in other subjects, you do need to have some form of trying to evaluate what the learner knows. But there again, my instinct was that based on my experience, say in France,

6:14where we had oral questions. In addition to writing essays, answers to questions on an exam, we also had this situation where we were confronted with a question, and we had to marshal our information about to think it through, we had to structure an answer in a period of 10 minutes, and then we had to speak for five or 10 minutes. And that kind of, which again, AI apparently calls regenerative testing, where you're forced to recreate based on all the things that you know, recreate an answer is more effective than short tests and quizzes, which are not natural and not as engaging for the learner. So here again, I think the AI, if it's done properly, can not only help us create these varied forms of the same content to enable us to learn these better and to explore them sometimes more difficult, sometimes less difficult, but also a format for demonstrating what we know by answering oral questions, which either could be put to us by a

7:17teacher, or could be put to us by an AI bot, and then we would have to record our answer. It may be that doing this in written format would also work, but I just wonder whether somehow the students would be able to cheat the system. Just as when we talk about creating different, you know, sort of more superficial or easy or summaries of certain material to hopefully induce the learner then to have or make it easier for the learner to to study more difficult material, if the learner stops with the summary and never does the more difficult reading, that of course will be very detrimental to their education. And there's already evidence that in universities, students are using this sort of some AI summaries, rather than doing the actual reading that's required, and the net result is that they do poorly on their exams. So I always feel if there's an opportunity for people to abuse a system, they will abuse a system. So it's very important that if we do use AI either to generate a variety of content to help us do this sort of circular

8:20learning as opposed to the traditional latter vertical learning approach, but also if we use AI as a means of testing learners on what they know in language or in other subjects, it's very important that we make sure this cannot be abused in some way. Anyway, those are just some of my random thoughts on AI and education and why I think this is an opportunity to rethink how a education is structured and how AI could enable us to overcome certain bottlenecks that may have existed before and which made the sort of traditional approach to education the only way to go. Now, I think with AI, we have a number of other pathways we can pursue to make education more effective. And in that regard, one of the things that is pointed out in the most beautiful history of intelligence by Duane and others is that the desire to learn things is essential to us as human beings. And there are rewards in the brain, triggering dopamine responses, a whole bunch of reasons why education is a fundamental desire of human beings. And we need to make

9:22sure that education, the way formal education is structured, actually encourages that basic desire in human beings and doesn't discourage it by introducing anxiety or discouraging exploration. Thank you. I look forward to your comments.