The fact that few students are self motivated enough to use it makes sense....but you are telling me that, in 4 years, _so_ few were motivated to use it that you can't report on whether or not it makes a difference for the minority that do?
I was among those who, when Khanmigo was first announced, were pretty excited about it's potential. I then waited for data on the results....and kept waiting.....and kept waiting. And now four years later this is apparently what we are going to get. I think that this is enough for me to decide that Khanmigo, regardless of whether or not a student actually engages with it, doesn't make much learning difference. At some point, the absence of (reported) data becomes data in itself.
I still believe, in principle, that AI tutors could be massively helpful for learning. But apparently we haven't yet figured out how to take that principle and turn it into reality.
I think the AI can not understand the situation of the student so it does not know what the student does not know. Therefore it can not guide the student through the main hurdles of learning and understanding a topic. Whereas a human tutor was once a human-student, AI never was.
I don't even think Khan Academy's original teaching revolution quite panned out.
I still remember when Khan Academy first came out, there was talk that teachers would go obsolete because teaching would become centralized and delivered over video.
Khan Academy to me is still just a YouTube channel trying very hard to be something more.
Indeed, reverse classroom, everybody getting access to high quality content, video then interaction, learning paths, etc.
Well, in practice it's still about the amount of time a pupil does train with the right oversight and that is precisely the bottleneck that hasn't been alleviated.
amazing in theory with the perfect user in the perfect use case,
misused in practice with terrible consequences for society at large.
Sure the one student who already excels, is motivated, understand what the concept to learn is, that actually completing exercises helps them to learn might, possibly, thrive. All other students, the vast majority, will try to "game" the (terrible) evaluation system to get good grades by cheating WHILE avoiding the very challenge that make the learning possible.
I think the teacher's job is not just to teach but more importantly to ingnite curiosity in students on new fields / subjects. If a genuine interest / curiosity in subject is missing then no amount or medium of tutoring can help.
AI is great for the curious. But its not yet there where it can proactively engage with students to generate interest.
You can be very AI-skeptic in various ways and still think that this is a fair take.
I teach and supervise students as master's level courses, and about 15% of them have intrinsic motivation to learn. These students have set up their own AI tutors with prompts and know way more than me in certain areas of the field, they are extremely ahead of their class.
The issue in my country is that you equate education with getting a safe job. 20 years ago, you needed a high-school degree in social science to get a government job. 10 years ago you needed a bachelor in social sciences to get the same job. 5 years ago you needed a bachelor in economy/engineering to get the same job. Now, because of recessions this is stretching to masters degrees.
You can't expect people who just want a job and a comfortable life and NEED to go to uni for this to want to be curious and want to learn.
> about 15% of them have intrinsic motivation to learn [...] they are extremely ahead of their class.
Feels like whatever tool they'd be given, they'd be ahead anyway. What's more worrying IMHO is, are the remaining 85% faring even worst than they would have before because they are learning even less, not just slower than the 15% learning faster. Namely is the gain for the few a loss for the majority?
And on other side education attainment has become metric for governments. More degrees and higher the degrees are better it will be for the economy somehow. Where there is likely quite a lot of jobs that don't actually need the degree.
This honestly seems like a flawed approach. Kids don't show up in the first grade, or the sixth, 9th, or really 12th, as the initiators of their educational journey.
An AI-based education system should have embedded in it "I am here to teach this person Geometry. Here is a list of the topics to cover, with a breakdown of steps for each including an intro section, a study section, a test section, and the meta material to go along with it.
I agree. I just can’t understand how people who work in education seem to be incapable of learning anything about education! They couldn’t have predicted that just having the most amazing tutor in the world available all the time just wouldn’t make any difference?? Students don’t go to school because they’re eager to learn. Don’t they know this? I think an effective tutor must be human or at least a human who can use the AI tutor on behalf of the child, which means the parent. Because your job as a tutor is not just spill knowledge, it’s to keep the student engaged ( just awake may be a challenge ) and make sure they are doing what they are supposed to.
Our kids elementary school (Germany) recently voiced concerns that pupils - especially boys - do not fulfill the self defined learning tasks in the expected time..
There may be no upper boundary to how much out of touch educators can be.
Might be harsh, but more I think about education more I think that the main job for many if not most kids or students in general is making them to learn. By various incentives be it punishments or rewards. Only really at highest levels is there any self-direction. And even there is plenty of external factors like not being paid anymore if you do not publish something.
Weird. No mention of the technical aspects, essentially just blaming average students for not being engaged enough with their simplistic ChatGPT clone. No wonder they have not dared yet to give out actual usage metrics. If I was given the choice between an inferior product that probably lags significantly behind on all features and one of the standard offerings from OpenAI, Google or Anthropic, I'd question why I should use this thing too. According to their website, they position Khanmigo like this:
>Unlike other AI tools such as ChatGPT, Khanmigo doesn’t just give answers. Instead, with limitless patience, it guides learners to find the answer themselves. In addition, Khanmigo is the only AI tool that is incorporated with Khan Academy’s world-class content library that covers math, humanities, coding, social studies, and more.
The first differentiation is literally just prompting (if at all). Nowadays you can tell any chatbot to behave that way. The second one may have been an edge before tool use was widely common, but with all chatbots now having access to the internet and code execution, it seems like this has also become a dud. This product was a nice idea on paper, but the fast technical evolution of the field has largely left it in the dust.
I mean, isn't the whole khan academy approach "we know better how to teach everything"? It's not surprising that they'd think they have more enlightened prompts than anyone else.
They had really cool math videos and got given too much money, that's about the story.
>Kristen DiCerbo, the organization’s chief learning officer, said AI can only respond to students based on what they ask. And it turns out, she said, “Students aren’t great at asking questions well.”
Ignoring whether or not this is a good idea in the first place, what about inverting the loop? Have the robot drive the interaction.
Yeah, that's weird. I've actually been working on an AI tutor for my kids that I'm thinking about open sourcing, but it drives the conversation in new directions using a concept graph that it can poll via tool use, and find the knowledge frontier for that learner.
It's been fascinating to watch - my kids are really into Slay the Spire, and it had a discussion about a decision tree they use when fighting one of the enemies, and then it used that to bridge to writing some python code and walking them through it. Another time, with dinosaurs, it went with them through the k-pg extinction event, and what really killed the dinosaurs - the kids thought the explosion - it walked them towards the sun dimming, and why food getting more scarce filtered for small mammals, our ancestors, and smaller dinosaurs.
on one hand I wish I would be smart enough to build something like that.
on the other hand, I was playing a lot Slay the Spire few years back and I would love to talk about with my kids while they play. Going from that it is not job of the parent to explain why dinosaurs are extinct?
How about completing the loop? Pose subject matter questions to them throughout the day, maybe via something like mobile push, collect their answers, immediately grade their results, and then actively reward them for performance.
All of the things brick and mortar schools are uniquely bad at.
>Khan Academy recently announced an overhaul of its product that provides students additional academic practice. Now Khanmigo is incorporated directly as a way students can get advice as they’re working through specific problems. A spokesperson said the organization made this change because “students were not seeking out Khanmigo’s help as much as we had hoped.”
Dear Lord, how is this any different from Microsoft sticking Copilot or Google sticking Gemini in every single offering? They're literally saying that people aren't using the chat bot enough so they're going to force it on people inside the product.
Well, there's a good way and a bad way to do this. Embedded AI (vs standalone AI) can do a lot of interesting things. For example, a good teacher will watch students and identify when someone is struggling (or thriving), and carefully offer to help, vs just waiting for students to approach them and aka questions.
There is an LLM button on every other website now. "Chat with your lesson", "chat with your food", "chat with your photos". People are not clicking them because they are just visual noise at this point.
It was honestly a very cool gimmick few years ago, when it was new. Chatting with your photos, imagine that! ... then after few tries you're like "well... I just want the last 10 photos because I want to share a photo of my son playing piano to his grand parents" and the UX, because it's actually well craft based on feedback from users, not only has a button for just that... but even default to this.
So... yeah, it got old quick. Genuinely cool for a bit but now "we" as users just want good UX. Now give me the FAQ that I can search through then an email if it's not in there.
PS: FWIW I do believe in a long-tail fashion, for few users who are not into scripting, might not be developers (or believe they could become) it could help find very few very niche use cases with solutions.
I'll probably open source and Show HN the AI tutor I've been working on for my kids at some point, but working on it has given me a little insight into the problem.
The biggest thing is motivation. First off, if Khanmigo requires them to type and read everything, that's going to get tiring fast for most kids. But I don't know how you could do voice in a school setting - mine uses STT/TTS, but with 20 kids in a room, it'd be chaos - STT accuracy and diarization with 2 is already really challenging.
Motivation is helped a bit by following their interest, but it seems like they're having trouble guiding the kids when they prompt it that way. That was a pretty big issue with mine early on - the kids would talk to it for an hour about whatever topic they were interested in at the time, but it would never branch into something new.
The tutor I'm working on solves it by having a concept graph that covers a lot of learning, from the basics like math, dinosaurs, etc to other developmental topics like 6 year old boundary-pushing humor, and two LLM threads - one that handles the conversational turns, but another one in the background that strategizes and steers thet conversational thread by looking at the concept graph connections and considering how ready they are for each, and then injecting steering notes into the conversational thread. Basically system 1 and system 2 thinking. And after sessions, it'll make a basic plan of where to start next time, and what might be interesting to offer up.
I mentioned this in another comment, but I've been really pleasantly surprised at the quality of the tutoring, especially when it bridges into new topics - one of my sons is really into slay the spire, and it's used that as a launching-off point into probabilities, decision trees, python code of the algorithms he thinks about as he's facing different enemies, and general strategies on different facets, and my other son was really into sharks, which it bridged into extinct sharks like megalodon, how scientists derive how it looks given cartilage's lower propensity to fossilize, bridging to dinosaurs and their fossils, the K-PG extinction event, how food scarcity filtered for smaller animals like the ancestors of birds, and our small mammalian ancestors. And a whole bunch of other topics.
It's been pretty great in that way, but my biggest open question at the moment is how to get them to engage with it on their own on a more regular basis - they go to it occasionally for random questions, but to get good coverage of that huge knowledge graph would take much more. And fundamentally, I think that human engagement still just has a number of important aspects to it that it's lacking, and I'm not sure if it's possible to replace those well enough to make it so kids want to learn a significant portion of their knowledge this way. And it's in their playroom, and a lot of times, they'd just rather play with their magnatiles than talk to this thing about how things work.
On one hand I will grow old knowing I'll always have a job because a lot of kids never will have researched anything in their lives and won't know how to deal with anything an LLM can't solve. On the other hand between this and most kids having had a 2 year covid gap in their learning, who the heck is going to pay my retirement and be my doctor when I'm old?
When I was young 20-30 years ago older people were saying that the Internet would make us dumb. Why learn anything when information is always readily available one search request away? Videogames were supposed to make me a blood-thirsty maniac, and don't even get me started on readily available porn. "Kids these days are lazy and don't want to learn" is one of the oldest memes in human history, with documented use going almost as far back as writing itself.
The flipside of that take is that if you listened to technologists, then educational TV/CD-ROMs/laptops/the internet/tablets/educational games/digital blackboards/MOOCs/etc. were going to completely revolutionize education - but looking at the evidence, it doesn't seem like students have gained much at all from any of it.
I remember an educator ranting to me a long time ago that the only data-proven ways to meaningfully improve educational outcomes was to reduce classroom size and make sure kids got enough sleep + fed well enough, everything else was just a waste of time.
There has been a measurable and noticeable drop in attainment starting with smartphones entering the classroom, supercharged by COVID chaos, and finally with AI cheating being just the latest assault on learning.
Ask teachers that have been teaching for 10 years. Ask the professors how today's kids are different than the ones of yesteryear.
The move to de-tech the classroom will eventually help out I expect, but keeping kids (and adults!!!) from using cognitive shortcuts so they can develop their own sense of what's reasonable instead of taking information from a bought-and-paid-for oracle is going to remain a problem.
I think it's particularly telling that the teacher dropped the product and the superintendent is saying it's going well. Maybe I'm biased from going to New Orleans public schools, but I have my doubts about how tapped in the superintendent is of the overall strategy.
That said I do think it's particularly hilarious that KA's strategy to students not wanting to use the product is to make the product more integral to the experience.
> “Students aren’t great at asking questions well.”
In my interactions with my kids public school and their teachers, they're goal is ram content down their throat and test for retention, not foster an environment open to questions
Had a teacher claim straight up they don't believe the system works and are just in teaching for benefits and summer vacation
IMO Sal Khan's revolution hasn't happened because the adults in charge right now are ignorant and inept but incredibly vain nonetheless
>In my interactions with my kids public school and their teachers, they're goal is ram content down their throat and test for retention, not foster an environment open to questions
Is that actually true though? Average American students (especially those in the public school system) are not excellent test takers, and they're even worse at rote memorization. If this is actually the goal they're not achieving that either.
Humans have 3 inch chimp brains. The expectations on those poor brains are too high. Its like expecting dogs to one day figure out how to run a mcdonalds. We do what we can. Theory of Bounded Rationality applies.
The vanity part has to do with Status Signaling. And status signaling is one of few know hacks that work at population scale in keeping all the different chimps together. Granted the status signalers some times get carried away and forget what its purpose is. See Theory of the Leisure Class.
I think it’s not right to blame teachers, because you cannot control that right? Agree it’s far too hard to find aspirational teachers, but they are naturally in demand and will be snatched up by well paying positions, like private schools. The tool is supposed to help given the current suboptimal teaching practices.
I was among those who, when Khanmigo was first announced, were pretty excited about it's potential. I then waited for data on the results....and kept waiting.....and kept waiting. And now four years later this is apparently what we are going to get. I think that this is enough for me to decide that Khanmigo, regardless of whether or not a student actually engages with it, doesn't make much learning difference. At some point, the absence of (reported) data becomes data in itself.
I still believe, in principle, that AI tutors could be massively helpful for learning. But apparently we haven't yet figured out how to take that principle and turn it into reality.
I still remember when Khan Academy first came out, there was talk that teachers would go obsolete because teaching would become centralized and delivered over video.
Khan Academy to me is still just a YouTube channel trying very hard to be something more.
Well, in practice it's still about the amount of time a pupil does train with the right oversight and that is precisely the bottleneck that hasn't been alleviated.
It’s going to be quite hard to motivate students to learn now that they know answering can be automated.
Who could have guessed.
AI is great for the curious. But its not yet there where it can proactively engage with students to generate interest.
The issue in my country is that you equate education with getting a safe job. 20 years ago, you needed a high-school degree in social science to get a government job. 10 years ago you needed a bachelor in social sciences to get the same job. 5 years ago you needed a bachelor in economy/engineering to get the same job. Now, because of recessions this is stretching to masters degrees.
You can't expect people who just want a job and a comfortable life and NEED to go to uni for this to want to be curious and want to learn.
Feels like whatever tool they'd be given, they'd be ahead anyway. What's more worrying IMHO is, are the remaining 85% faring even worst than they would have before because they are learning even less, not just slower than the 15% learning faster. Namely is the gain for the few a loss for the majority?
An AI-based education system should have embedded in it "I am here to teach this person Geometry. Here is a list of the topics to cover, with a breakdown of steps for each including an intro section, a study section, a test section, and the meta material to go along with it.
That would work.
The people who work in education don't have this issue; the people who work in tech and assume that gives them expertise in education do.
>Unlike other AI tools such as ChatGPT, Khanmigo doesn’t just give answers. Instead, with limitless patience, it guides learners to find the answer themselves. In addition, Khanmigo is the only AI tool that is incorporated with Khan Academy’s world-class content library that covers math, humanities, coding, social studies, and more.
The first differentiation is literally just prompting (if at all). Nowadays you can tell any chatbot to behave that way. The second one may have been an edge before tool use was widely common, but with all chatbots now having access to the internet and code execution, it seems like this has also become a dud. This product was a nice idea on paper, but the fast technical evolution of the field has largely left it in the dust.
They had really cool math videos and got given too much money, that's about the story.
Ignoring whether or not this is a good idea in the first place, what about inverting the loop? Have the robot drive the interaction.
It's been fascinating to watch - my kids are really into Slay the Spire, and it had a discussion about a decision tree they use when fighting one of the enemies, and then it used that to bridge to writing some python code and walking them through it. Another time, with dinosaurs, it went with them through the k-pg extinction event, and what really killed the dinosaurs - the kids thought the explosion - it walked them towards the sun dimming, and why food getting more scarce filtered for small mammals, our ancestors, and smaller dinosaurs.
on the other hand, I was playing a lot Slay the Spire few years back and I would love to talk about with my kids while they play. Going from that it is not job of the parent to explain why dinosaurs are extinct?
How about completing the loop? Pose subject matter questions to them throughout the day, maybe via something like mobile push, collect their answers, immediately grade their results, and then actively reward them for performance.
All of the things brick and mortar schools are uniquely bad at.
If you can’t articulate what you want it becomes a guessing game
Dear Lord, how is this any different from Microsoft sticking Copilot or Google sticking Gemini in every single offering? They're literally saying that people aren't using the chat bot enough so they're going to force it on people inside the product.
So... yeah, it got old quick. Genuinely cool for a bit but now "we" as users just want good UX. Now give me the FAQ that I can search through then an email if it's not in there.
PS: FWIW I do believe in a long-tail fashion, for few users who are not into scripting, might not be developers (or believe they could become) it could help find very few very niche use cases with solutions.
That is a warning sign if ever there was one.
The biggest thing is motivation. First off, if Khanmigo requires them to type and read everything, that's going to get tiring fast for most kids. But I don't know how you could do voice in a school setting - mine uses STT/TTS, but with 20 kids in a room, it'd be chaos - STT accuracy and diarization with 2 is already really challenging.
Motivation is helped a bit by following their interest, but it seems like they're having trouble guiding the kids when they prompt it that way. That was a pretty big issue with mine early on - the kids would talk to it for an hour about whatever topic they were interested in at the time, but it would never branch into something new.
The tutor I'm working on solves it by having a concept graph that covers a lot of learning, from the basics like math, dinosaurs, etc to other developmental topics like 6 year old boundary-pushing humor, and two LLM threads - one that handles the conversational turns, but another one in the background that strategizes and steers thet conversational thread by looking at the concept graph connections and considering how ready they are for each, and then injecting steering notes into the conversational thread. Basically system 1 and system 2 thinking. And after sessions, it'll make a basic plan of where to start next time, and what might be interesting to offer up.
I mentioned this in another comment, but I've been really pleasantly surprised at the quality of the tutoring, especially when it bridges into new topics - one of my sons is really into slay the spire, and it's used that as a launching-off point into probabilities, decision trees, python code of the algorithms he thinks about as he's facing different enemies, and general strategies on different facets, and my other son was really into sharks, which it bridged into extinct sharks like megalodon, how scientists derive how it looks given cartilage's lower propensity to fossilize, bridging to dinosaurs and their fossils, the K-PG extinction event, how food scarcity filtered for smaller animals like the ancestors of birds, and our small mammalian ancestors. And a whole bunch of other topics.
It's been pretty great in that way, but my biggest open question at the moment is how to get them to engage with it on their own on a more regular basis - they go to it occasionally for random questions, but to get good coverage of that huge knowledge graph would take much more. And fundamentally, I think that human engagement still just has a number of important aspects to it that it's lacking, and I'm not sure if it's possible to replace those well enough to make it so kids want to learn a significant portion of their knowledge this way. And it's in their playroom, and a lot of times, they'd just rather play with their magnatiles than talk to this thing about how things work.
I remember an educator ranting to me a long time ago that the only data-proven ways to meaningfully improve educational outcomes was to reduce classroom size and make sure kids got enough sleep + fed well enough, everything else was just a waste of time.
Ask teachers that have been teaching for 10 years. Ask the professors how today's kids are different than the ones of yesteryear.
The move to de-tech the classroom will eventually help out I expect, but keeping kids (and adults!!!) from using cognitive shortcuts so they can develop their own sense of what's reasonable instead of taking information from a bought-and-paid-for oracle is going to remain a problem.
That said I do think it's particularly hilarious that KA's strategy to students not wanting to use the product is to make the product more integral to the experience.
> “Students aren’t great at asking questions well.”
In my interactions with my kids public school and their teachers, they're goal is ram content down their throat and test for retention, not foster an environment open to questions
Had a teacher claim straight up they don't believe the system works and are just in teaching for benefits and summer vacation
IMO Sal Khan's revolution hasn't happened because the adults in charge right now are ignorant and inept but incredibly vain nonetheless
Is that actually true though? Average American students (especially those in the public school system) are not excellent test takers, and they're even worse at rote memorization. If this is actually the goal they're not achieving that either.
Who would have thought?