Try a geometry problem with AI. Either is uses latex notation that doesn’t make sense without a Latex interpreter (which isn’t included by default), or it gets it wrong. As someone who teaches remotely to high school students, most chat applications aren’t set up for Maths properly as yet, so the notation we use to type with doesn’t match what we hand write. How we express Maths problems is still a larger issue with technology.
Absolutely yes. My suspicion is this is a solvable problem but if it isn't solved on a five year timeline I think all of my bets default to my favor. EdLight is an interesting group to watch here. Pedagogically and technologically serious IMO.
In teaching, relationships matter; a community of learners matters; talking through what been learned matters; equal access matters. I can imagine where AI could "improve" getting a "quick" answer. But for the hard, fun and exciting work of actual teaching and learning, I am not predicting a sea change in those critical areas with AI.
It's pretty interesting to put this conversation about teacher attrition and job satisfaction with this study I saw the other day about how teacher interest, prestige, and satisfaction have changed over the past 50 years (through presumably a great deal of technological advancement): https://www.edworkingpapers.com/ai22-679
Someone I was talking to at a party this past weekend asked me the same question ("Do you think AI will replace teachers?"), and my answer was similar. I think it will help improve instruction and learning, but things will remain mostly unchanged.
The areas I thought could change significantly, though, were hybrid learning and college intro courses.
1.) There may be more offerings for kids who don't need as much help that are AI-led with a teacher in the periphery for them. But at that point, I'm not sure that's a teacher or if that's an AI or IT expert.
2.) College intro courses are already crammed with as many students as possible. I could see some colleges trying to make these even more efficient with AI.
But both of these are predicated on AI becoming more accurate.
> College intro courses are already crammed with as many students as possible. I could see some colleges trying to make these even more efficient with AI.
Agreed. In situations where students lack a relationship or responsiveness from their teacher, AI will perform closer to the mean, and maybe exceed it.
Veritasium has an entertaining video about how Radio was meant to revolutionize education, then it was TV, the videodisks, tablets... https://www.youtube.com/watch?v=GEmuEWjHr5c
Yeah, I'm not aware of its replication. I wouldn't personally set it as the bar for success for any of my work, but then I'm not one of the AI chatbot boosters who constantly hold themselves to Bloom's standard. They chose the terms of the wager, is what I'm saying, not me.
A reader proposes a bet that is bet tougher on my case and I'd still take it without hesitation:
> In the next five years, we won’t see a peer-reviewed RCT where students receive the same dosage of either a chatbot tutor and a consistent human tutor and the chatbot tutor group significantly outperforms the human tutor group for academic achievement.
Dan, while I agree with your overall worldview about AI's place in education, I do think you have overlooked an important part of the argument with the following bet:
"In the next five years, we won’t see a peer-reviewed RCT where students receive the same dosage of either a chatbot tutor and a consistent human tutor and the chatbot tutor group significantly outperforms the human tutor group for academic achievement."
This seems to be a bit of a red herring to me. Let's assume for simplicity that there are three options:
1. No tutor.
2. AI chatbot tutor.
3. Human tutor.
Further, let's pretend that an AI tutor is 30% as effective as a human tutor. In this scenario, you've clearly won your bet due to the underwhelming efficacy of the AI chatbot, but you have overlooked the positive impact of such a chatbot. There are a finite number of human tutors and a finite amount of money to pay tutors. If AI chatbots allow students who would otherwise have no tutor get a tutor (albeit a 70% less effective tutor), I think this is likely a net "win."
Now, does that "revolutionize" education? Of course not. But, I think as you draw conclusions and form opinions about AI, you can't leave out AI's important advantage--scalability.
I don't disagree with a word here. I'd love it if the AI chat boosters would quantify the kind of effect they anticipate on a five year timeline. What is this revolution we should anticipate? That'd give us something to talk about. Sal Khan has, himself, made the comparison to Bloom so that's what I'm taking as the terms of this wager. But there are outcomes that are far south of revolutionary that are still really good. Like .5 SD for example. I would take the under on that bet as well FWIW.
Thank you for this article; I think it is one of the few articles on AI in education that made me think. And the reason is that it didn't totally fit in to my understanding, or my approach, so to speak, to AI and education.
Last thing first; I agree with you, AI won't replace human teachers until, as you pointed out, people will want it or won't care. Still, even then, I believe there will be crazy people like you, not only the author of this very article but an educator who impacted many educators all over the world, and us, not only the readers of this very article but a bunch of those educators who have been being impacted by you, that will still become teachers and perform their profession with love and passion. This profession won't die out for sure.
I am with you on the belief (unless proven or disproven by a research, of which you bet won't happen, it's a belief right?) that kids won't prefer AI over us, humans. I think, no, I know that a healthy, professional, positive teacher-student relationship helps their learning significantly --both from my experience and from research results, as well as Rita Pierson's TED talk --one of the few that moves me, together with yours.
The best quote I have ever heard about AI replacing anyone in any profession (not only teachers) is that, "AI won't replace you, but someone who knows how to use AI, will!" In his recent article, Dan Fitzpatrick wrote that the days where students brought their phones to schools would be a nice nostalgia, as each student would have a "personal AI" in their pockets. I don't see this impossible; after all, I was quite happy playing Snake on my Nokia until I bought a new model in awe, on which I can save a single-note ringtone of Smoke on the Water, like 15 years ago. I still remember how hard it was for me to comprehend that I could sign out from my Facebook account on the school PC and sign back in at home, and everything was exactly as it was when I signed out!. And now, we are talking over AI which is already within reach for many uses.
Hence, no, everything is about to change drastically. Education will follow at a slower pace and it's not going to happen everywhere at the same pace, like it hasn't been since, but it will.
I am unsure about your third bet, but I am certain that you will win the first two bets; there won't be any national survey or peer-reviewed study to prove them wrong. Because these methods will be obsolete, too. No one will care if masses think this way or that, or how well a certain method work in an experiment group compared to a control group. If there is one thing that will change in education, it is that it will be personalised. Hence it will matter whether John has been progressing in his learning more than a 1000-student experiment group's whatever statistical values show significant positive change in their learning compared to the control group. In other words, Qualitative Research model(s) will be dominant; this is my bet.
#1 Can you point me to research that says a tutor who is right is more important that a tutor who is curious and stokes curiosity in the tutee?
I'm just thinking about Sugata Mitras "granny cloud" and don't know if this idea went anywhere. I hope were not just using AI as a solutions guide/answer sheet.
#2 I have yet to see mass adoption of the one to one tutor to pupil suggested by Bloom (I'm not well read and do not recall this outside of reading your post). I think that providing students with a highly encrypted AI life coach (tutor and councilor in school) is a terrifying yet worthy route to explore. Though it is probably not going to happen in 5 years (as a linear thinker). If we can somehow get this "tutor" to everyone, I see that as an equitable endeavor.
The other question to ask, is even if AI chatbots are better at teaching do we want our children raised in a society where they look to AI chatbots as teachers. No one thinks it is a good idea to raise our children among wolves, do we really want them to be raised among "robots"? (Apologies for any grammar mistakes, AI chatbots were not used.)
Try a geometry problem with AI. Either is uses latex notation that doesn’t make sense without a Latex interpreter (which isn’t included by default), or it gets it wrong. As someone who teaches remotely to high school students, most chat applications aren’t set up for Maths properly as yet, so the notation we use to type with doesn’t match what we hand write. How we express Maths problems is still a larger issue with technology.
Absolutely yes. My suspicion is this is a solvable problem but if it isn't solved on a five year timeline I think all of my bets default to my favor. EdLight is an interesting group to watch here. Pedagogically and technologically serious IMO.
https://www.edlight.com/
In teaching, relationships matter; a community of learners matters; talking through what been learned matters; equal access matters. I can imagine where AI could "improve" getting a "quick" answer. But for the hard, fun and exciting work of actual teaching and learning, I am not predicting a sea change in those critical areas with AI.
It's pretty interesting to put this conversation about teacher attrition and job satisfaction with this study I saw the other day about how teacher interest, prestige, and satisfaction have changed over the past 50 years (through presumably a great deal of technological advancement): https://www.edworkingpapers.com/ai22-679
Super helpful citation, thx.
Someone I was talking to at a party this past weekend asked me the same question ("Do you think AI will replace teachers?"), and my answer was similar. I think it will help improve instruction and learning, but things will remain mostly unchanged.
The areas I thought could change significantly, though, were hybrid learning and college intro courses.
1.) There may be more offerings for kids who don't need as much help that are AI-led with a teacher in the periphery for them. But at that point, I'm not sure that's a teacher or if that's an AI or IT expert.
2.) College intro courses are already crammed with as many students as possible. I could see some colleges trying to make these even more efficient with AI.
But both of these are predicated on AI becoming more accurate.
> College intro courses are already crammed with as many students as possible. I could see some colleges trying to make these even more efficient with AI.
Agreed. In situations where students lack a relationship or responsiveness from their teacher, AI will perform closer to the mean, and maybe exceed it.
Veritasium has an entertaining video about how Radio was meant to revolutionize education, then it was TV, the videodisks, tablets... https://www.youtube.com/watch?v=GEmuEWjHr5c
The two-sigma study is such a pipe dream. Searching on https://www.evidenceforessa.org/programs/math/ or https://ies.ed.gov/ncee/wwc/FWW/Results?filters=,Math the interventions with tutors are not even close to one sigma. What am I missing? Sounds like a fraudulent study.
Also improving school performance two-sigma would be equivalent to raising 30 points of IQ, wouldn't it?.
> The two-sigma study is such a pipe dream.
Yeah, I'm not aware of its replication. I wouldn't personally set it as the bar for success for any of my work, but then I'm not one of the AI chatbot boosters who constantly hold themselves to Bloom's standard. They chose the terms of the wager, is what I'm saying, not me.
A reader proposes a bet that is bet tougher on my case and I'd still take it without hesitation:
> In the next five years, we won’t see a peer-reviewed RCT where students receive the same dosage of either a chatbot tutor and a consistent human tutor and the chatbot tutor group significantly outperforms the human tutor group for academic achievement.
don't forget crayons, the revolutionary invention of the first decade of the 20th century.
Love Veritasium!
I was also concerned about this two sigma standard.
Which district has access to 1-1 tutoring for all of its students?
I haven't seen that as an option in my neighborhood.
Peanut Gallery Request: What is your percentage confidence in each of your bets? That's what the "superforecasters" do. (https://en.wikipedia.org/wiki/Superforecasting:_The_Art_and_Science_of_Prediction)
Dan, while I agree with your overall worldview about AI's place in education, I do think you have overlooked an important part of the argument with the following bet:
"In the next five years, we won’t see a peer-reviewed RCT where students receive the same dosage of either a chatbot tutor and a consistent human tutor and the chatbot tutor group significantly outperforms the human tutor group for academic achievement."
This seems to be a bit of a red herring to me. Let's assume for simplicity that there are three options:
1. No tutor.
2. AI chatbot tutor.
3. Human tutor.
Further, let's pretend that an AI tutor is 30% as effective as a human tutor. In this scenario, you've clearly won your bet due to the underwhelming efficacy of the AI chatbot, but you have overlooked the positive impact of such a chatbot. There are a finite number of human tutors and a finite amount of money to pay tutors. If AI chatbots allow students who would otherwise have no tutor get a tutor (albeit a 70% less effective tutor), I think this is likely a net "win."
Now, does that "revolutionize" education? Of course not. But, I think as you draw conclusions and form opinions about AI, you can't leave out AI's important advantage--scalability.
I don't disagree with a word here. I'd love it if the AI chat boosters would quantify the kind of effect they anticipate on a five year timeline. What is this revolution we should anticipate? That'd give us something to talk about. Sal Khan has, himself, made the comparison to Bloom so that's what I'm taking as the terms of this wager. But there are outcomes that are far south of revolutionary that are still really good. Like .5 SD for example. I would take the under on that bet as well FWIW.
Thank you for this article; I think it is one of the few articles on AI in education that made me think. And the reason is that it didn't totally fit in to my understanding, or my approach, so to speak, to AI and education.
Last thing first; I agree with you, AI won't replace human teachers until, as you pointed out, people will want it or won't care. Still, even then, I believe there will be crazy people like you, not only the author of this very article but an educator who impacted many educators all over the world, and us, not only the readers of this very article but a bunch of those educators who have been being impacted by you, that will still become teachers and perform their profession with love and passion. This profession won't die out for sure.
I am with you on the belief (unless proven or disproven by a research, of which you bet won't happen, it's a belief right?) that kids won't prefer AI over us, humans. I think, no, I know that a healthy, professional, positive teacher-student relationship helps their learning significantly --both from my experience and from research results, as well as Rita Pierson's TED talk --one of the few that moves me, together with yours.
The best quote I have ever heard about AI replacing anyone in any profession (not only teachers) is that, "AI won't replace you, but someone who knows how to use AI, will!" In his recent article, Dan Fitzpatrick wrote that the days where students brought their phones to schools would be a nice nostalgia, as each student would have a "personal AI" in their pockets. I don't see this impossible; after all, I was quite happy playing Snake on my Nokia until I bought a new model in awe, on which I can save a single-note ringtone of Smoke on the Water, like 15 years ago. I still remember how hard it was for me to comprehend that I could sign out from my Facebook account on the school PC and sign back in at home, and everything was exactly as it was when I signed out!. And now, we are talking over AI which is already within reach for many uses.
Hence, no, everything is about to change drastically. Education will follow at a slower pace and it's not going to happen everywhere at the same pace, like it hasn't been since, but it will.
I am unsure about your third bet, but I am certain that you will win the first two bets; there won't be any national survey or peer-reviewed study to prove them wrong. Because these methods will be obsolete, too. No one will care if masses think this way or that, or how well a certain method work in an experiment group compared to a control group. If there is one thing that will change in education, it is that it will be personalised. Hence it will matter whether John has been progressing in his learning more than a 1000-student experiment group's whatever statistical values show significant positive change in their learning compared to the control group. In other words, Qualitative Research model(s) will be dominant; this is my bet.
I do not disagree with your bet's, but...
#1 Can you point me to research that says a tutor who is right is more important that a tutor who is curious and stokes curiosity in the tutee?
I'm just thinking about Sugata Mitras "granny cloud" and don't know if this idea went anywhere. I hope were not just using AI as a solutions guide/answer sheet.
#2 I have yet to see mass adoption of the one to one tutor to pupil suggested by Bloom (I'm not well read and do not recall this outside of reading your post). I think that providing students with a highly encrypted AI life coach (tutor and councilor in school) is a terrifying yet worthy route to explore. Though it is probably not going to happen in 5 years (as a linear thinker). If we can somehow get this "tutor" to everyone, I see that as an equitable endeavor.
The other question to ask, is even if AI chatbots are better at teaching do we want our children raised in a society where they look to AI chatbots as teachers. No one thinks it is a good idea to raise our children among wolves, do we really want them to be raised among "robots"? (Apologies for any grammar mistakes, AI chatbots were not used.)