AI in Education After 4 Years: Tutor Revolution or Learning Crisis?

0:00Four years after Gen AI entered the mainstream public life, we hear two radically different predictions about education. First is optimistic, even revolutionary. I will give every student a personal tutor, adapt lessons to individual needs, provide immediate feedback, make learning, more engaging and reduce teachers workload. Second prediction is much darker. I will weaken attention, replace independent thinking, make assessment unreliable, precise isolation, gradually remove teachers from the centre of education. So after four years, which of these has been proven to be correct? Welcome back to the teacher's AI cafe. I'm Kane, a teacher and a parent who wants AI to give us back our time. I share practical classroom tested AI strategies to cut down your workload and prove your feedback and reduce stress so you can get your evenings back and be more presently a friend of family. No jargon, just actual steps, grab your mug, and let's get going. So we're going to compare two different perspectives. The first comes from Catherine.

1:02I think it's Burberry, Burberry, our Sing, a British head teacher known for strong views on discipline, knowledge and teacher authority. She's known as the most strictest headmistress in the UK or England at least. She's quite popular when you listen to her talk, she's definitely worth listening to. The second actually comes from MIT's 2026 report on artificial intelligence in teaching, engineering research training. Catherine's primary view is concerned with schools, MIT's examining university education. But despite these two different settings, they share more in common than we may expect. So part one, the original promise, that I started again, part one, the original promise. Machachipity first appeared in 2022. Same to offering some extraordinary educational opportunity, imagine personal tutors, students can ask the same question repeatedly without being embarrassed, an explanation could be simplified, expanded, translated. Feedback would arrive immediately, rather than days later.

2:02Teachers could use AI to prepare lesson plans. Create resources. Do those admin work? The university students could use it to analyze data and prove code, explore complex texts, undertake projects, and would not normally have been possible within a semester. The promise has not entirely disappeared. How I can generally access or improve access to AI explanations and feedback. It can help students begin a task when they feel stuck and support people's learning in the second language, in those living with disabilities. It can also help teachers create resources, spend less time on repetitive tasks, and that's really what my channel's about. Before using experience, I expose the weakness in the original vision. Education is not simply the transfer of information from knowledge. System to a student. And a personal tutor is not very more mellow because it's actually available at that point. Students must still choose to use it, concentrate, persevere and care about the actual learning. This is where Katherine begins her criticism. So part two, students do not work for screens.

3:05That is Katherine's argument. And it is very valid. And people often talk about the computer games, but that's not the same as education. Let's see. Katherine rejects the idea that education is simply a process of download and knowledge. And she addresses women at all. When she, what I bought off was, you'll talk about Eva and I'm not talking about, is simply learning the algorithm and embedding that and understanding it, then your success or which is there's more to education than that. She argues that children do not automatically select the activity that produces the longest, with greatest long term benefit. When given a choice between difficult academic work and easy entertainment, many would understand, well, you choose the easier option. And this is where she also argues about this point since the gamification of education. Now, it doesn't really solve the problems we're trying to solve. Turns of effort and perseverance with the students. Now, personalization does not solve this problem. Now, our tutor can adapt an explanation, but it can not fully reproduce the relationship with when a teacher can develop.

4:07Where they develop the authority, no seers of students behave yet, and demand greater effort from that student. Teachers see their hesitation, distraction, embarrassment, and avoidance. Teachers can say, I know you can do better than this. That actually has a big impact. More importantly, the statement matches because it comes from another person. Catherine argues that children often worked because they respect their teachers. They want to make their parents proud. They do not want to disappoint the people believing them, which is the truth. Students or kids won't learn from teachers they don't like. From this perspective, education is not only about knowledge, it's also about character. Students learn how to concentrate, respond to correction, overcome frustration, continue working when success is not immediate. Now, if AR removes every difficulty, it may also remove the experience that the perseverance actually develops in the students. Our central message is pretty simple.

5:08Schools do not need machines that replace teachers. They need capable teachers who use technologies carefully and remain responsive for the classroom. I can't disagree with that in any shape or for. Part 3. MIT discovers the university version of the same problem. MIT reports begins with more acceptable position, more of an accepting position. I do suggest to go read the report it is public-lovable. It's not a huge report. Not much to what's common sense, but quite acceptable address. A recognized AI is already embedded in education and professional life. University students will need to know how to use it effectively, responsibly and ethically. However, MIT identifies many of the same risks described by Catherine. Students can now use AI to produce essays, solve problems, write proof generic code. That creates an assessment crisis. When students submit a post assignment instructing my non-longer

6:09no, whether it's actually the student really understands it. And there's obviously one going. There's many ways around that, particularly at the high school level. It's very easy. But MIT argues that the problem is deeper than cheating. The greatest dangers that many students may successfully complete the coursework without completing the intellectual work that the coursework was developed to produce. Like said, we do everything with a purpose, don't we? We don't just hand out things we'll leave in. We'll be nearly to keep them busy. If someone gets finished but the learning does not happen, they won't do it. T calls missing ingredient productives. Struggle. Now, say to my students, your brains got to hurt. You got to grow it. Learning requires a degree of cognitive friction. That's it, brains hurting. It's growing, developing your straniate. Students need to wrestle with difficult problems, make mistakes, revise their ideas, and gradually develop understanding. A.O. I can support this process, or allow students to bypass it completely. MIT also report also signs that A.O. is changing social life

7:09over education. Students may be attending fuel office hospitists, stating lesson discussion, replacing study groups with prior conversation with chatbots. A.O. makes students feel supported, while simultaneously making them feel more isolated. It's where the school and the university perspectives meet. Now I agree with them on that before I come on this point that often the greatest learning that happens at university, in my opinion, obviously, you go to the content, is the discussions you have with the people you're attending with and the ideas sharing and the debates that you have. Either supports or picks holes in your arguments and you actually get better from it. Also, that social networking development will benefit you when you move in the workplace because networking is everything for jobs. Anyway, I digress in that point. So Catherine argues that children learn through relationships, authority and accountability. MIT says that university students learn through instructors, teaching assistants, study groups, laboratories, mentors and research communities. But they're both really the same conclusion,

8:10even if they call them different things. Education is a social process, not million information service, and I agree with that 100% particularly in the secondary. The high school area. So part four, school and university are not the same. Now, obviously there are important differences between the two different settings. The school students are still developing, foundational knowledge, sustained attention, software regulation, independent judgment. They might know enough to recognise and pay our answers wrong, that's very important. This creates a serious paradox. So to use A.O intelligently, a student already needs the knowledge to actually start with, and emphasises that they need their knowledge to ask the right questions, and it will assess what comes out. They need enough subject to understand to ask a meaningful question, evaluate the response and identify a fabricated claim, to determine whether the reasoning makes sense. We cannot replace fundamental knowledge with A.O. skills, because fundamental knowledge is what makes responsible A.O. use possible.

9:14For younger students, this support a really cautious approach. Skills need to protect significant periods of A.O. free reading, writing, calculation, discussion, and memory work, and this is what the data is starting to show, what's getting implementation. Linked into that reduced screen time for different reasons. University students can reasonably be given great freedom, as they're preferring to work in professional, where A.O. is, in many cases, going to be very common. The grading for him does not eliminate the need for boundaries for that. Even Harley-Couple University students face workload pressure. A.O. offers a quick route to better grade the temptation to substitute completion for learning reminds me of. That's what we do. We always find the shortest way to get something done. A perfect role of A.O. would therefore change the student's age, knowledge, maturity, and learning objectives. So the question is not really the should A.O. use A.O. Should students use A.O.

10:14The better questions are, what must students first learn without? When does A.O. support the learning process, and when does it perform the learning task, or another student's behalf? So let's go and look at part five. What is the evidence shown in these four years? Now, the evidence is mixed. And I see these person, Facebook, they put so many reports out, constantly on A.O. I think they're just tuning them out. I don't know their background or from their good and Indian name, that doesn't mean us. But like I said, they're obviously, they're absolutely prolific. Does make me question the actual basis of the research, these papers they're putting out. But, international research supports the concern about digital destruction, and that's more of a general of the whole screens thing. Across the board, I've talked about that a few times, and there's lots of media around that. OECD findings shows an association between frequent classroom destruction from digital devices, and lower academic performance. But the same way of it, it's also suggest that post-full

11:15directed technology use can be beneficial, which has always been the case. That, that distinction is very important, because it's the teacher's direction with a specific goal in mind, instead of just the device being there all the time. So smart phone use spent a time is not the same as the educational software, selected by the teacher. An evidence about tablet screen time and older, educational technology is not automatically proof that AI harms learning. They're very different. Actually, I've been reading a listen to podcasts, it's a listen to good books could kill. And what they do is they find very popular books, and they actually research them and go into the background. And it's funny how one's a recent listen to, they pulled apart the anxiety generation thing was called, and they just pulled apart the basis and the research of this book. And that's why when we look at things talk about that and on the background of the basis research, it does point out that we can get the wrong conclusions

12:16if the research and the data really poor and written in a way that doesn't truly tell us what it did, but it's really swathing someone's agenda. So once again, as educators, we need to be careful of that. And it's not the reason why a lot of fads come along. And also why it takes time to process. And that's why I think four years in, we're still learning about this process. But let's get back to the main point. But so when we look at different claims, we've got to be careful with the dramatic claims, because they're there for a reason. Falling test scores after introduction of devices, dust and us doesn't necessarily prove the devices along course that decline. Social conditions, teaching methods, curriculum changes, penting, disruption, can also result in affect results. Likewise, high usage does not prove educational success. Students who use and I actually, chewed a frequently without developing independent mastery, will save a lot of time, the same time in the not gaining the actual knowledge.

13:17Engagements not the same as understanding otherwise, we could just make amazing amounts of computer games that subconsciously teach in all these skills and knowledge. Computer games teach different sorts of skills. Very valuable ones in my opinion. So producing correct answers not the same as being able to explain, reproduce, the reasoning later on. My important measure should be the transfer of knowledge after receiving help from IOA. Can the students solve a new problem independently of getting that help? If not the system may have improved the task completion, but there's been no improved learning. Part six. What is education doing now? So government institutions begin to move away from the simple choice between banning AI and embracing it. In Australia, we've got a national framework for GNIO in schools, focus on responsible and ethical use, human wellbeing, transparency, fairness, privacy, security and accountability. At no time, or place does it actually mention specific types of AI.

14:17England's Department of Education says AI may help teachers, particularly with planning and admin work, but should not replace professional judgment or a teacher, student relationship. Ginesco recommends human centered, an Asia-Propiate approach, which is particularly attention to children's privacy and safety. AI tutoring companies are also learning that implementation requires teachers. Providing software is not enough. Schools need professional development, clear routines and ways for teachers to monitor with the students of learning. At uni level, MIT recommends redesigning courses rather than attempting to put an AI doesn't exist. It proposes clear course policies or examinations, portfolios, project work, presentations, and assignments followed by in-person discussion. It recommends structured social learning, strong mentorship, AI literacy, privacy protections, transparent disclosure when instructors use AI themselves. Because I think there's a bit of media around that, if they're feeling ripped off, told not to do it, and they know their lecturers and instructors are actually using it.

15:20So emerging repotters that, I guess, neither teacher, I guess, neither teacher replacement or complete technological rejection. What we want is in a teacher led integration process. And when we talk about teacher led, that means every single school, it's potentially going to look very different. Every network or district, other than you, like US terms, are going to look different, because all our students are different. Hunts 7, practical role for AI. And a lot of these I would have repeated, or if you'd have sent me for a while. So what should AI's role being in education after four years? First, it can support teachers, most 100%. They can listen to what podcasts can do. Trafture, first draft of your course materials, adapt reading levels, generate practice question and reduce admin work. The teachers must verify the results

16:21and remain professional responsible for it. Second, AI can produce practice and feedback. A world-us-on system might ask, guiding questions, identify misconceptions or hints. Should resist immediately supply lead answer? It can translate, explain how familiar a vocab provide alternatives. Fourth, it can expand or advance work. Once students possess sufficient and find their actual knowledge, AI can actually help them. Fifth, AI literacy should be part of education. Students need to understand the hallucinations of the biased, the privacy, authorship, intellectual property, and the environmental cost of our ship reducing it. Though that's not as dramatic as some people say, in some contexts. They must know when to use AI when not to use it. AI cannot become a primary source of the motivation of guidance, discipline, or being in human connection. And it should not be the sole evaluator of students work.

17:22It should not replace teachers for responsibility. It should not remove the independent practice from quite the build fundational knowledge. So, part eight, are responsible for a test for responsible use. Perhaps a simple way of testing and introducing AI is into any education activity with us, maybe four questions. What is the student meant to learn? Which part is the necessary thinking that the AI will perform? Can the student later demonstrate the school with that AI? Does this use strength and will work in the relationship between the student, the teacher, and the learning community? Very simple questions, but, like I said, very powerful. But, like I said, the teacher controlled. Now, if you can answer those questions, and you're getting the yes and four yeses, yep, implement it. Now, feeling yes, and then knows, if they, for example, can't demonstrate the school with that AI, no, that isn't a good use of it. You need to do something else. So, while we're looking at move,

18:23making sure we don't move that thinking, accountability, and human interaction that we set the core of our educational system. As a matter of how flawed it may be, it is the best one that we actually have at the moment. Now, the distinction is that it's about augmentation, and not automating the thinking process. I guess that's probably one of the more available lessons for the first four years. So, let's wrap this up. So, after four years, AI has not made teachers obsolete. And I gave a recent talk to all her teachers. To them, AI isn't coming for their job. More importantly, actually revealed how you can play down the standard of teaching sometimes. What? Teachers do more than deliver information. They establish expectations. They model judgment, they recognize confusion, they create belonging, they challenge students to persist, and they help people understand why learning matters. They also contribute to education, but not carry its full human purpose. The future should not be a contest between teachers' machines. The real task is to place AI within an educational system.

19:28One which teaches retain the authority, students the responsibility, and the technology is judged whether it produces something valuable. The question is no longer whether AI will be present in education because it is here. The question is whether education will shape AI's role, or AI will shape education according to the values we never consciously chose to do. So, what do you think? Can AI should have developed independence? Or does it risk creating dependence? Which past the education should always remain human? And how should we know if AI was helping students, rather than simply helping them finish?