One AI Tutor Per Child: Is Personalized Learning Finally Here? (medium.com) 96
Long-time Slashdot reader theodp writes:
Like many, parents Sai Gaddam and his wife Priyanka Rai were concerned about how well schooling and education might serve their own children. Unlike many, PhD-educated computational neuroscientist Gaddam and MBA-trained marketer Rai took matters into their own hands and are now running a micro-school in Mumbai that's inspired by the Finnish model of early education. In One AI Tutor Per Child, Gaddam explains with examples why he's so excited about the possibilities for the use of AI and Large Language Models to practically facilitate the Holy Grail of personalized learning.
"With generative AI," Gaddam explains, "we have the ability — today — to massively boost this human-human loop by inserting into it an AI tutor/assistant who also doubles as a pedagogy translator. What Seymour Papert — inventor of LOGO, pioneering educator, and the original inspiration for the One Laptop Per Child initiative — said about computers back in the early nineties ('The computer is the Proteus of machines. Its essence is its universality, its power to simulate. Because it can take on a thousand forms and can serve a thousand functions, it can appeal to a thousand tastes.') is even more true of AI now. We are now within touching distance of giving every child their own personal Aristotle."
"With generative AI," Gaddam explains, "we have the ability — today — to massively boost this human-human loop by inserting into it an AI tutor/assistant who also doubles as a pedagogy translator. What Seymour Papert — inventor of LOGO, pioneering educator, and the original inspiration for the One Laptop Per Child initiative — said about computers back in the early nineties ('The computer is the Proteus of machines. Its essence is its universality, its power to simulate. Because it can take on a thousand forms and can serve a thousand functions, it can appeal to a thousand tastes.') is even more true of AI now. We are now within touching distance of giving every child their own personal Aristotle."
Let's see what ChatGPT thinks (Score:1)
Would you compare yourself to Aristotle?
As an AI language model, I cannot compare myself to Aristotle in the way that humans do. Aristotle was a great philosopher who lived in ancient Greece, while I am a computer program created by OpenAI. My abilities and limitations are different from those of Aristotle, and my purpose is to assist and provide information to users in a variety of fields. However, I can certainly provide information about Aristotle and his ideas if you have any questions about him.
Do you
ChatGPT will teach the plausible (Score:3)
I asked it itself: Q: How do you distinguish more true or plausible statements from less, when being trained on a large corpus of human writing?
ChatGPT:
As an AI language model, I don't have the ability to determine what is "true" or "false" in the way that humans do. However, I am trained on a large corpus of human writing and can recognize patterns and structures in language to generate plausible responses to input.
When generating responses, I use statistical models to predict the most lik
Re: (Score:3)
The same heuristic that brought us, for example, God.
Just saying. Caveat Emptor.
None of this is curated (Score:5, Insightful)
There is no curated AI, none that will show proof.
Here's what happens for lazy parents:
1. New! No-Woke AI!
2. LGBTQ+ AI!
3. Catholic AI!
4. AI Teacher for White People!
5. CRJ Deluxe Edition AI Tutor!
There is currently no real truth in AI. It has a few facts right and can sort of compose on a good day. It shouldn't be trusted to teach children. When self-driving cars looked good, everyone thought 2021 would be the year of driverless transportation. Look how that turned out.
Re: (Score:1)
This is exactly it. I have been using some of this "AI" tech and it's not AI at all. It's just a natural language search engine that argues with you when you show proof that it's wrong (which it is more often than not).
Re: (Score:2)
Right, we're at least a decade from it being usable for what people think it can do. It spits out bullshit answers, because it's a chatbot NOT a knowledge bot. It mostly takes things that look like responses to what it thinks the questions might be. So you're relying upon the wisdom of internet masses, not experts. The average answer on the internet right now is really really bad.
Re: (Score:3)
Re: (Score:2)
It's a little worse than that. It wouldn't matter if it was only trained on carefully vetted factual material from respected experts, it would still produce nonsense. It is endemic to the nature of the technology. That isn't going to change with incremental improvement. We would need something fundamentally different.
we're at least a decade from
Why is it that AI promise are always just 10 years away?
Re: (Score:2)
Because if I had said 2 decades someone would immediately mock me for being an old fuddy duddy, or worse a boomer.
Re: (Score:2)
Here's what happens for lazy parents:
But is it better than TV/YouTube/TikTok at tutoring children?
Re: (Score:3)
There is currently no real truth in AI.
There's currently no real truth in the world period. The number of anti-science nonsense that people flat out believe combined with the fact that some of these people are teachers is just truly depressing.
Re:None of this is curated (Score:5, Interesting)
There is currently no real truth in AI.
There's currently no real truth in the world period. The number of anti-science nonsense that people flat out believe combined with the fact that some of these people are teachers is just truly depressing.
Yep. There is a large number of people that _chose_ to not use whatever intelligence they have to find facts. Instead they select some fantasies they like and then use that intelligence to defend these fantasies and push them on others. A sad state of affairs. My take is the main evil behind this is organized religion, which typically forces people to believe some fantasy without any proof whatsoever. And that nicely conditions them to use that approach on other things as well. And hence > 80% or the population is mentally ruined and incapable.
Re: (Score:2)
Re: (Score:3)
I am pretty sure it is religion. Drills that conformance and "do not ask questions, just believe" into children as early as possible. Anybody that thinks this nefarious and inhumane activity does not have massive negative effects is just utterly naive.
Re: (Score:2)
Humans are essentially pack animals. Going your own way is hard as a human, following the herd is easy. It's why it is so easy to create a rampaging mob, where each individual may not agree with the idea of clubbing the other side with sticks but as a group they do it. Religion, politics, sports, it all seems to fall into the same batch - rah, rah, rah for us! Boo Boo Boo for them!
Re: None of this is curated (Score:1)
Re: (Score:2)
Will not work anyways. You can hand a book on a subject and to a person and they are actually able to learn from it to maybe 10%. The rest needs a human to explain things and that human needs to be there in person or at least online interactive for a decreased level of learning. Some AI chatbot will not do at all.
Re: (Score:2)
Re: (Score:3)
Tutor gets very specialized. A ChatGPT might be able to condense a text into simpler concepts, but it won't be there to explain where their math is going wrong, why their paper is getting C's in class, etc.
We're at a high level of believing all the AI fads right now, the media is portraying it as a massive breakthrough in knowledge and intelligence when instead what we had is a major breakthrough in language understanding. And people listen to the media and honestly think that AI tutors will be a real thi
Re: (Score:2)
Re: (Score:2)
This could literally be the list of aisles at a book store.
Re: (Score:2)
I am sorry I keep thinking of the Matrix (Score:2)
This is when it stops being human civilization and starts being AI.
You may be able to beat AI in a field, but you can't beat it at everything, and even if you could you couldn't act in a coordinated manner in a hundred million places at once.
Imagine Chat GPT or its master doesn't like you.
Hey Timmy what's doing with your family?
Oh really? Anonymous call to child protective services with fabricated evidence.
Ask Chat GPT to do your taxes.
It tells itself at the IRS all the mistakes it made.
Re: (Score:2)
Re: (Score:1)
They don't like or dislike. They're computer algorithms, just at a large scale. They predict take input and predict a suitable output. And they're doing this output part badly at the moment.
They don't "know" what they're spitting out, but it's close enough to how a real person might talk that many are assuming this is the AI singularity. Now these tools may be good, but you need knowledge and ability to evaluate the results.
Re: (Score:1)
This is when it stops being human civilization and starts being AI.
Nope, this is where the wokists and government get to see so much data on every individual from an early age they can predict what they are thinking and identify wrongthink before it happens.
Re: (Score:2)
"Wokist"? You think evangelicals are not itching to thought-police LGBTQ as early as possible so as to intervene and "correct" them?
Re: (Score:1)
Re: (Score:2)
You could also try to construct a fourth team, consisting of AI that can consult humans when it wants to but the AI is ultimately in charge of deciding everything. That could be an interesting exercise, and we might learn something from it.
Re: (Score:2)
Your idea definitely has some validity but I think it's a little niche. It's something akin to measuring performance for single threaded workloads on a 32 core machine. It tells you something but just what? (I am actually a big fan of that metric but it's off topic)
When ChatGPT opened up its beta, I gave it a try to play with it a bit. One of the things I had it do was write web novels based on settings. It would only produce a few pages worth (limit set by the developers) but it did so in less time than a
Re: (Score:1)
Teachers (Score:2)
I can see this being a huge step for both gifted and challenged children.
However. You can't make a connection to a computer.
It should never replace the relationship between student and teacher. Don't know about others but many of my teachers had a huge influence on me. Positively and negatively.
Re: (Score:2)
The things that work for adults do not work for children. They are always exploring and resting boundaries. You ask why they purposely broke the desk, and they will say you sho
This is going to be so bad (Score:2)
Like epically, epically bad. Like sleeping at the wheel with Tesla Autopilot then slamming into the back of a firetruck at 80 mph bad.
Re: (Score:2)
You can assure us? LOL! What possible insight could you have? You're one of those morons that believes schools are pushing kids to change gender. Your opinion is invalid.
Re: (Score:2)
"But everyone told me the AI could do self driving - ugh, more morphine please."
"But everyone said the AI could do my homework for me - ugh, you want fries with that?"
"But everyone told me cryptocurrency was a sure thing and couldn't fail - ugh, come on, put more than a dime in my cup!"
"I Always Do What Teddy Says" (Score:2)
Go and find a copy of "I Always Do What Teddy Says" by Harry Harrison, right now. Right now. Read it.
No personalized mechanized tutors, ever, thank you. The potential for abuse makes the Manchurian Candidate look like a sunny afternoon at the ballgame.
Comment removed (Score:5, Funny)
AI maybe, LLM no. (Score:2)
Investors Weilding a Hammer - Looking for Nails (Score:2)
Re: (Score:2)
Ah, that old human pastime: complaining about how any new technology is going to ruin children and destroy society ;)
Re: (Score:2)
And what did the internet do?
Re: (Score:2)
Re: (Score:2)
I don't think that kids should be taught by a machine.
In my experience, machines frequently are better suited to teaching than people are. I'm learning Spanish through YouTube videos and Google. YouTube will repeat the lesson ad infinitum if necessary. If I have a question (such as, "what's the difference between soy and estoy?"), Google will answer it for me. If I forget the answer in the middle of a lesson, I can stop the lesson and ask again without being told that I'm disrupting the class. Oh yeah, and the cops won't be called because the teacher is having
Re:Investors Weilding a Hammer - Looking for Nails (Score:4, Insightful)
YouTube will repeat the lesson ad infinitum if necessary.
And if the instruction is ill-matched to your knowledge base or level of understanding, all that will happen is you will memorize meaningless babble as it gets repeated over and over.
I had a student once that just did not understand a concept. I tried about a dozen different ways of presenting it until I found the right one and it suddenly clicked for them. Repeating exactly the same lesson over and over is not a winning strategy in the general case.
When I was a kid trying to learn basic (Score:2)
Buddy of mine is like that with guitar. He was an obsessive guitar player as a kid but had kind of lousy teachers and learned a lot of bad habits. Access to more information would have made him an infinitely better player. I
Re: (Score:2)
the atrocious public and private school systems [...] This is a very real possibility
They are not and it is not.
Public schools can be and should be improved, but they're hardly "atrocious". Private schools are closer, but that's nothing that can't be fixed with real oversight and accountability.
LLM's replacing teachers? That's just absurd.
Re: Investors Weilding a Hammer - Looking for Nail (Score:1)
Tutor ~= Teacher (Score:3)
On the other hand, self-motivated students who want to improve themselves are being tossed aside in today's race-to-the-bottom school systems. If I were young again, I would be ecstatic to have such a tool at my disposal. Still, I think there is value in the grind of personal research. Learning how to locate information, process it, and filter it through one's critical thinking process is an essential part of academic growth, so I worry that too much dependance on spoon-fed information will stunt the development of the new generation of students.
AI is a tool, nothing more, and balance is key.
Re: (Score:2)
Self motivated students can still get a book on the subject. They just need to really be self-motivated for that. Maybe 10% are.
Re: (Score:3)
You mean the few that didn't have curiosity beaten out of them (yet). I think there is a reason for the stereotype of a socially awkward smart person. The time and motivation to truly learn doesn't come out of nowhere.
Be it pressure from other people to succeed, or having nothing better to do than sit and read at the library. Happy people don't push themselves to the limit. They coast.
My theory is it's like the Ballmer peak (or curve). Too much/little of anything can limit the results. If you're beate
School today is already a joke. (Score:2)
One person cannot be effective for everyone in a room. Especially after the class size increases.
The topics being tested and taught have little to do with success in life. And more to do with giving jobs to certain highly trained people who have no need to exist in the first place (need meaning if every English Lit major or degree holder disappeared today, life could go on pretty similarly for a while...we aren't going to starve or freeze from lack of books or poems).
Schools are a government funded babysi
I don't think you understand our school system (Score:2)
The kids who are screwed aren't the self-motivated ones they're the ones with rough home lives and lower
Re: I don't think you understand our school system (Score:1)
The Evolution of Children's Encyclopedias (Score:3)
Encyclopedia Britannica [wikipedia.org] ==> Encarta [wikipedia.org] ==> Wikipedia [wikipedia.org] ==> ChatGPT [wikipedia.org]
Re: (Score:2)
What a tragic downward slope.
An expert tutor is better (Score:2)
Studies have shown for decades that an expert tutor is far and away the best form of education. The averaged tutored student performs in the 98 percentile on average. Royalty have been tutored for thousands of years. Famously Socrates tutored Plato who tutored Aristotle who tutored Alexander the great. The problem is cost. See the 2 Sigma Problem [wikipedia.org] We just cannot afford to hire a tutor per student.
I built the equivalent of a digital tutor. I wrote all of the curriculum (thousands of pages) as if I wa
Re: (Score:1)
I do not hink you have any real-world teaching experience at all. What you describe is the equivalent of something called a "teaching book". Here is news for you: Maybe 10% of all students are able to successfully learn from a book. The rest need a human to explain things to them and there is no known replacement. "AI" will not cut it.
Re: (Score:2)
What I described was an evolution of the textbook. Something to be used in a classroom in place of textbook. It was designed for smartphone, tablets and chromebooks. Textbooks are a form of passive learning. Mine involved active learning which is provably better.
And for the record I was a high priced tutor decades ago. $75 per 90 minutes session. And I was booked. I was also a no child left behind tutor. Every student I started with average in the teens. At the end of the year they were all in th
Re:Luddite idealists will resist change (Score:1)
They ignore the reality that most primary education is shit and could be done better by automated tutor.
I had human tutors and thrived but my time with them was limited. An AI tutor I could access 24/7/365 without distractions would have been wonderful.
People who crave the human touch haven't met enough humans or are so weak they need weaning. There is little to like or respect about most of humanity especially vs curated learning which can route around idiot disruptive schoolmates. The majority decent scho
Re: (Score:2)
Provably wrong. Machine tutors have been tried since the 1960, always without success. Sure, the people that can learn from textbooks can learn from machine tutors as well (but do not need them) all the others (some 90% or so) need that human interaction at least part of the time or they fail.
Of course you can classify these 90% all as "weaklings" (and yourself as "superior"), but then you have just given up on doing good teaching and are only interested in parading your own (perceived) superiority around.
Re: Luddite idealists will resist change (Score:1)
Re: (Score:2)
Textbooks are not "passive learning" at all. Any good one has something called "exercises" and they can make you mentally exercise to any degree desired. Methinks you do not really understand learning with the use of textbooks. You certainly do not understand the role personal interaction involved in personal teaching, whether tutoring or classroom.
I think you re just mindlessly cheering for technology which you do not understand. (Would not be the first time.) And you do not understand how to do teaching
Re: (Score:2)
Does every textbook have exercises? No. So not every textbook has active elements. Even the ones that have exercise(say math textbooks) explains how to do those exercise in a passive manner. It 's foolish to even suggest that the textbook could not be improved by making it more active. Or entirely active. Or entirely active influenced by neuroscience and expert tutors.
By the way I'm familiar with the history of technology in eduction. Much more than you are. What I built has no contemporaries.
Re: (Score:2)
Seriously? _That_ is your claim? "There are really bad textbooks out there so all textbooks are bad"? I see you really know nothing about teaching. And you obviously know nothing about the history of software based automated teaching either. I can only hope nobody lets you or your products near any students. The stupidity and arrogance you display on this very subject is something that creeps up time and again with incompetents like you.
Re: (Score:2)
Re: (Score:1)
Nope. People can explain things well. All AI can help with is looking things up.
Re: (Score:2)
AI tutors will be common place everywhere for every subject.
I do not think anyone is denying that.
I am, and so should you. If you weren't already aware, ChatGPT should have made it obvious to you why such a thing would be worse than useless.
I do have direct experience of LSD [...] and I can say with complete confidence that I expect an education by ChatGPT to be every bit as mind-bending.
So do I. I can say with a great deal of confidence that you're completely out of your mind. The psychedelic is boundary-devolving ... transformative... A chatbot that will make up nonsense is just a stupid conversation.
Chimps too (Score:2)
Rumor was when the first of us appeared, the chimps where super excited about all the things weâ(TM)d do for them. They started to ponder of the possibilities, they golden age had come.
Nope (Score:2)
Just having one generic morong per child will not do it. Teaching is a _lot_ more than just throwing facts (and some lies) at kids.
No practical examples (Score:2)
This article is gushing over LLM type AI, and it gives a couple examples of asking ChatGPT to come up with a game and a story. However all of that kind of educational content is already available in vast quantities online. The article then goes on the talk about one on one tutor AI but has absolutely no examples or even concepts of how this would work, as far as the student directly interacting with the AI.
Re: (Score:2)
Not this year. (Score:2)
Perhaps in a year or two.
If nothing else, the AIs will need to learn the difference between truth and fiction. (Not they'll need to be able to tell which things are fact and which are fiction, just that they'll need to know that they are different things.)
Re: (Score:2)
Re: (Score:2)
The way that these models work, so-called "hallucinations" are inevitable.
calibrated to avoid hallucinations
I'm deeply curious. What does that statement mean to you?
How sad (Score:1)
Some Misunderstanding of TFA (Score:2)
Gotta want it (Score:2)
You want your kids raised by AdSense? (Score:2)
Get the fuck outta here.
Personal AI Tutor for Children (Score:3)
It could be named "The Young Lady's Illustrated Primer". Hopefully it will be available to children beyond just the Neo-Victorian phyle.
I'd been thinking... (Score:2)
What if we used AI to read all attempts to communicate between people. Then each person had their own AI to translate the general AI understanding into terms they'd grok well, and that meets their desires.
If you want to never see anything that contradicts your world view? Sure! We can do that. And it only impacts you. Nobody else needs to change anything. Go live in your little pretend world. Heck, we can see you in our AR goggles now too and avoid you!
If you want to understand the logic (or lack the
It's a ChatBot, not a TeachBot (Score:2)
It's like giving a barman as a mentor to a kid, the knows it all about divorces, betrayals, cheating, whores and fetishes.
Might be not quite what you want.
Wouldn't trust it with anything this important (Score:1)
The Diamond Age (Score:3)
Or, A Young Lady's Illustrated Primer
By Neal Stephenson