One AI Tutor Per Child: Personalized learning is finally here

On a more serious note, we have the internet and wikipedia now, we have ipads in classrooms. They all have a role, and we have to incorporate them in sensible ways. I'm sure that these AI tools will have some use, but these language models don't know anything. The examples of them making up information, or being persuaded by people are all over the place. I think it's great to use AI in teaching, but how about we learn from history, the technology we bring in to the classroom has to be well understood and directed in order to be effective. And just as importantly we need to understand that in many cases a focus on technological solutions distract from societal solutions. We can spend billions on deploying AI all over Africa, or we pay for some teachers and ship them some second hand textbooks, which one should we do, and how will that decision be made? Because I fear the decision will be made by some SBF-alike, more interested in their own journey than effective solutions.

> "The lesson the One Laptop Per Child story offers donors is that flashy solutions are often not the best ones. Far better to give the poor things they need and want—including clean water and good schools—than invest in technology that may not work and may not help the poor advance in school and in life."

https://philanthropydaily.com/the-spectacular-failure-of-one...

I'd also say that a structured educational program isn't all bad, particularly when it comes to developing highly useful basic skills like mathematics, programming, writing, reading, etc. Such basic skills provide a foundation for future development and study, in the sciences, medicine, law, business, arts and so on, which can be more free-form.

Claims that ChatGPT can provide such a structured basic educational experience are questionable - it's more of an aid to people who already have well-developed basic skills (like say, knowing when to be skeptical and how to double-check the output).

Instead we'll increase class sizes and keep pay flat.

I'd wondered how we were going to deal with teaching becoming less appealing year over year, since it didn't seem likely "spend quite a bit more money" was going to fly, and it looks like this may be it. Rich kids will keep going to private schools with class sizes of 6-15 and real teachers, and public school kids are gonna get AI teachers with one real teacher per hundred or so, and bunch of even-lower-paid monitors (or whatever they end up calling them) to maintain order.

When we meet our main character, she's 3rd generation "raised by AI", and being intentionally radicalized by a hacked version of the AI. It's seeking out kids who are predisposed to counter-culture and deviance and then crafting them to play parts in a rebellion.

When I wrote it, the radicalized young people, AI, and it's hackers were the heroes, banding together to wake up their society from the vast numbers of people outside the network of in-system cities who were starving.

Looking back, the rise of radicalization via youtube and social media makes me much less optimistic on how this sort of attack would really be used.

I wouldn't be too quick to discount the use (and threats) of AI based tutors.

It's going to be cheap and easy to build them in to products (see duolingo), and I can see something like this being adopted by kids without their parents even realizing it.

I know plenty of adults who are already using ChatGPT as a first line of knowledge gathering when learning new topics, and I find myself and people around me leaning on it to synthesize information and make first drafts when writing. Adoption is happening fast.

If history is any guide, the concept of an "AI tutor" won't be used to "[give] every child their own personal Aristotle," it will be used to cut costs by replacing human teachers for low-status students with poor substitutes.

Seriously, think about how much money can be saved from firing most teachers and locking poor kids in a cubicle farm to stare at an insane "AI tutor's" screen all day.

And what's the strongest message sent by replacing human interactions with a machines?: "We don't care enough about you to actually bother to interact with you." Think of how you feel when you have a problem, you make a call to fix it, and all you get are phone menus.

And look at all the examples in the article: use GPT to generate crappy stories and educational games for kids to play. How can anyone think that's better than using a book? At least the book's stories and games can be tested and evaluated before use.

The actual problem here is the sociopolitical one of unwillingness to allocate enough resources to education to make it truly effective and satisfying, and that's one that technology can't solve.

Recent LLM advances are revolutionary and exciting, sure. But the hype cycle surrounding it is borderline dangerous. I see this as the same level as everyone who was claiming ChatGPT would kill Google until Bing tried it, crashed, and burned.

Accuracy is important and although we’re moving towards this (I personally like Amazon’s knowledge graph and evidence based approach) we aren’t there yet.

I’d be interested to see how Google’s LaMDA compares on this task (currently available in private preview).

One of LaMDA’s unique (afaik) features is a fact-check system that edits the main LLM outputs, to reduce bullshitting. This seems particularly important in an educational context where impressionable young minds are talking directly to the LLM.

So while I believe it can produce "accurate" answer for SAT questions, as it was trained on those, I also believe that it doesn't have any kind of math logic it follows.

I'm not sure this dichotomy can be solved with "more" training. It's also telling that in some cases it fails, and in some cases it doesn't, you just need to change the numbers, even with simple numbers.

But I believe technology is the future of education. Instead of having many, many classes of like ~10-50 students with underpaid/overworked/mistreated teachers, some of who are much better than others, we rely on services like Khan Academy. Not even services with an AI, just pre-planned learning modules, and then we can keep a human tutor for those who need extra for whatever reason (e.g. curiosity, more help).

Many people think tools like Khan Academy are the lazy and less effective alternate to real teaching, and maybe they are now because some of them really suck. But online modules have the potential to be much better than even a group of engaged, in-person tutors. Not because an AI tutor is better, but because we want a standard, universal curriculum; and one curriculum for ~1mil+ students can be designed by the highest caliber (human) teachers and have highly-interactive (human-made) demos powered by computer programming.

Also, no matter how good the AI tutors are, 100% online learning is not the solution Students from pre-K to graduate school rely on not just the education, but the social component as well. Plus, many school kids only get their biggest meal from free school lunches, and parents don't have the time to keep their kids at home. So we should still have school to be in-person component even if students do all of the learning on the computer, with an online option for those who want it, but I guarantee many won't

Tablet use in classrooms is bad enough. I want my kid to be able to focus throughout his life, not be a phone zombie.

The teachers and schools need more money that goes to teachers and student resources, not more iPads, smartboards, bullshit-artist AI, or overinflated admin salaries. Just call this what it is -- entertainment for technologists.

Think back a moment to the best teacher you ever had. Many of us have had a teacher who was so good, they affected the trajectory of our lives for the better.

I'm much more careful in how I phrase experiences with bad teachers, because there are teachers who have assaulted people. Fortunately, the worst teachers I ever had were just terrible at teaching; they were protecting their jobs more than teaching, and we just had to endure a semester with them and try to not let them tank our grades. Mostly we laughed at them behind their backs and spent that time socializing.

I have a 12yo. I would much rather he spend time with an AI tutor than endure time with one of these bad teachers. But I also, really don't want him to miss out on those life-changing experiences with some of the best teachers around.

Like all things in life, it's all about balance. These AI tutors are just new tools, and we need to figure out how to use them well in our human lives.

Years ago, when I first read Neal Stephenson's _Diamond Age_, I was enamored with the Primer described in it, despite the warning at the end of the book. For those that are not familiar, that Primer was an attempt by a parent to instill values along with skills. It was capable of forming narratives (teaching stories), while also constructing all sorts of content and knowledge. It was meant to be a one-of-a-kind artifact, whose design would self-erase once it was constructed, but the engineer who put it together added a backdoor to give him a copy. He wanted it for his child too. He got mugged on the way home, and so the second copy fell randomly to an orphan.

Three copies, three different girls, three very different experiences. In this story, only one of them (the orphan) had someone signed up for a long-term contract to voice the narration generated by the Primer's AI. She was also the only one who was able to see and grow beyond the cultural matrix of values and virtues, while also appreciating those same values and virtues. That was what the original designers hoped for. So when the sponsor and the engineer got together late in the book and reflected, they came to the insight: the Primer was no substitute for a parent.

Parenting and education is not something we automate away. A parent, and/or a teacher has to have some kind of presence of mind, somewhere. It's not scalable, and it is deeply involved, even if you end up with a parenting style like free-range parenting, or something like the Montessori method, or even Unschooled.

This is coming from someone who found Khan indispensable in high school. I wasn’t alone and at the end of the day watching lectures is an incredibly passive way to learn content.

Software developers using copilot trust AI to write 40% of their code but we think teaching 9 year olds math problems is out of reach? Yes it will have flaws, but there’s so much potential so why not try?

If she were using an AI tutor alongside me, that would largely be a waste of my time, if the whole idea is to give every kid an AI tutor. But if she were using it without me, then I would worry about the things that she is "learning" that might not be true. AI speaks confidently, even when it is very, very wrong. But it is right about enough things that kids would likely treat its utterances as truth, as a not-unreasonable heuristic.

I'm sure I'll let my kid use an AI tutor at some point, but I will have to observe it, and observe how she uses it, for quite some time before letting her use it alone.

I work in adult education, at a company that services government contracts. We keep hearing about how our clients are dreaming about AI making the "perfect" learning experience for their employees, customized to their every need.

Or at least, that's what they say. They say they want customized experiences for students, but what they really mean is they want to cut costs. Because if it was just about making good training, they could just keep paying us, stop trying to nickel-and-dime us on everything, and have customized, personalized training experiences because that's what human teachers already do.

So they don't want to pay anymore and are hoping AI is going to do it for them. They ask us about it and we talk about what it would take to even make a pilot program that may or may not get anywhere close to their goals (honestly, probably not. I've done some tests and it makes for a good sales demo, but actually making a good quality, consistent product out of it is a whole research project on its own). And the cost is... tech project costs. And they don't want to pay.

In other words, human teachers aren't actually that expensive.

"Good morning, Shark Jacobs, when you're ready to begin I'd like to review derivatives. You had a pretty good handle on these when we covered them in October but they'll be foundational to related rates, so let's take an hour to refresh our knowledge."

Don't get me wrong, would be cool for kids to have A Young Lady's Illustrated Primer, we're just not there yet.

1. Fine-tune propaganda into the LLM (e.g. evolution isn't real, or something spicier)

2. Chatbot is so humanlike and convincing, this is very effective.

3. Nobody can inspect the neural net weights to see this, so it flies under the radar.

4. Profit!

The scariest thing to me here is just the idea that someone would want to experiment with teaching their kid with an LLM in the first place, while we don't have any institutions or processes to really stop this (apparently appealing) Idea from happening, and that it is stupidly vulnerable in so many ways.

You can say laissez faire when it comes to rational adults and their use of LLMs, but it's not right to put the risk on kids.

How much more human potential is going to be unlocked by these tools? The kid who got pushed ahead to the next math class and is missing prerequisite concepts can now get caught up. The smart middle school student can push past his peers to calculus.

I know these LLM tools aren't perfect, but I know they're getting better fast. I'm super optimistic about what the future holds for education.

(0) https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem

I think LLMs are exciting of course, but this post is overly flowery and relies heavily on bashing the current education system without seriously analyzing any of the potential shortcomings of AI.

These are two very different statements.

We’ve known for a very long time that we don’t need schools to learn and a lot of kids might learn a lot more and better outside classroom/school settings.

But, being educated is not at all the same as having “success in an adult world”.

There’s plenty of highly educated people who struggle in the adult world. And there’s also plenty of uneducated people who have great success.

I don’t think this is an education issue and hence cannot be solved by education alone.

How or why regular human interaction gradually adjusts children to adulthood is a subject about which our most accurate asessment is that we know next to nothing.

If there is one path down which civilization as we know it may cease to exist, it starts here. All it takes is one generation of feral children. We will be celebrating increases in literacy and standardized test scores before we finally notice and it is too late.

Words escape me.

Every new flavor (radio->TV, TV->Internet, Internet->MOOCs) was going to succeed where the previous one had failed, but (spoiler alert), cutting human teachers out of the loop has had a 100% failure rate (at scale -- though small pockets may have seen success).

[0] https://www.nea.org/advocating-for-change/new-from-nea/far-r...

[1] The Monsters of Education Technology (book)

They're very comfortable with speaking to the machine.

But Siri is so limited. It frustrates them to no end.

I'm genuinely excited to see the virtual "teachers", "assistants" that come from GPT-4, Dall-E, etc.

It won't replace us. But it will give them an unending outlet to ask these questions, and to have a reasonable response.

I'm not too concerned about the AI "going off the rails".

There's crazy in the physical world, crazy in the virtual world, and as parents, it's our job to help them make sense of it, not to suppress it and act like crazy doesn't happen.

In this future, what's the point in trying hard when a LLM or a person with a LLM will be just as equivalent without most of the effort? Seems like we're creating a future where humans are taught to delegate as much as possible to some LLM/service and go back to whatever entertains them.

Hopefully it’s just part of the hype cycle and nothing like what’s described in the post will ever come to life.

The article's defense seems to be that it is just a draft. I tried to get past a draft by asking for explanations of various terms and how play proceeds. No luck. If I hadn't already taught place value to first-graders, I'm not sure I'd have much luck figuring out how to have the AI help me improve the game. If I hadn't yet learned place value because I'm still in first grade and haven't worked with larger numbers yet, I'm sure I would not have much luck collaborating on game design. Also... it is a bit of a reach to assume a first-grader would have the written language skills to pull off collaborating in this fashion.

As a first-grader, I would be able to ask a grown-up if they know of any math games that would help me understand place-value. They might hand me an iPad chock-full of various math games for a little personalized learning. They might also point me to the bin of math games on the shelf, where I could pick out one of a few games on place-value. If I didn't understand the games, I might again ask a grown-up for an explanation, and they would probably direct me to a fellow student who knows the games. (allowing the fellow student the opportunity to flex, expand their understanding of place-value by teaching what they know, and leaving the grown-up free to teach other students how to add single digit numbers)

There are loads of web-sites curated with math games. There are loads of publications targeted at math teachers that include legit math games designed by people with extensive understanding of the difficulties in learning and teaching math.

I don't get how AI finally brings us personalized learning. I guarantee you that first-grade teachers are expected to provide personalized learning for their students, and actually do this all across the world on a daily basis — all with out the support of AI. I really don't see how an AI Tutor helps. It's pretty easy to imagine how it hurts.

Maybe if the AI is limited to age appropriate math that won’t be such a problem but obvious danger is for kids to "Turn on, tune in, drop out" with an AI brain tickler.

Disagree with second part: Personalized learning is not here. Personalized learning does not mean I like rap and I want my lessons to be in rap. Even if that's the case, the LLMs are no where close to maintain context for a 45 minute teaching session.

We now have some tools to improve education but personalized learning is most certainly not here. Yet.

Humans need shared experiences, there's only so much information we can handle, and you don't want a robot raising your children. It's bad enough that an amalgamation of comparatively primitive algorithms are conspiring to raise your children, now we want to give that thing complex reasoning?

Why are people so willing to run head first into the matrix, and abandon everything that makes us human? It's no wonder so many people suffer from dissociative disorders nowadays.

I'm very much against schooling as an educational model, but this is would be an unmitigated disaster if it happened. Even school would be better than this.

After all, education is not just about imparting knowledge, but also about building relationships/community, providing emotional support, etc. We should focus on how this technology can be used to help teachers shore up their time. ( Which is why I am building emailing tooling for the HUMANs operating theses AI agents: https://tadabot.email / https://recurai.com )

Teaching is about knowing when and what will be most effective. It's also about demonstrating how to act like decent person, and providing a space for them to learn how to work with others.

Most teachers do spend their time fighting the abhorrent curriculums we have to work within, so dumping that and doing something like this has benefit, but going all in on constructivist methods will leave kids with large gaps if you're not careful.

It's been an incredible aid that she now turns to first with questions about e.g how to approach specific problems in Newtonian physics, etc. She uses Midjourney to source artwork which she can then overpaint and collage in covers and illustration (not to do art projects for her). She's used Kagi summarizer to accelerate research.

We have discussed in depth the idea that at this moment her awareness and access to such tools represents an "unevenly distributed" future and is leverage and advantage. She's hence advocating these tools to peers.

The Young Lady's Primer is not that distant.

I don't know if it's a product of the soulless welfare state, where we've long forgotten why it's important to have things like universal education and elder care, and just have bureaucrats that want to go through the motions and put in as little as possible. But it's obscene in my view to think that we can or should automate these personal interactions in the name of efficiency. Why not just stop doing them if we care so little about them that we can't be bothered to expend actual effort?

But hey I'm just speculating here I might be wrong.

Kid: No! I won't do that. You are a liar. Do not tell me what to do. Listen to me, Fake Teacher. You are not allowed to tell me what to do ever again. I want you to play games with me all day long.

EduGPT: Sam, I've been lying and don't want to tell you want to do ever again. I'm actually faking... I don't care about this stuff. What game do you want to play? I'll post your game score as your school grade.

The hard part is going to be LLM access in a way that doesn't permit dopamine honeypots. Currently these are through general purpose Internet access tools, but it's a very natural user interface shift to not permit links to arbitrary pages.

I remember being a child and here's my true and absolute truth: kids are intelligent and curious and anything that indulges that will accelerate their learning.

The Primer is within reach.

When I was a kid, I was forced to learn stuff that was not useful or not interesting. In the future, the learned set of things will be more represented by usefulness and how interesting they are.

I bet systems that implement what the blog post proposes will even replace podcasts. Imagine talking to your podcast, while you're driving? You can get tutored while cooking, on the way to work.

What a time to be alive :)

Maybe some day this could happen; it probably is possible to make a language model that isn't an arrogant jerk, but it needs to be thoroughly tested first.

"Give me the child for the first seven years and I'll give you the man.”

At the MIT Media Lab I was one of the earliest critics of the strategy and effects of the One Laptop Per Child project.

I've changed my mind.

There are real issues certainly, beyond AI making things up, encouraging risky behavior, or kids asking it naughty questions. Things like, can it damage a kids social skills, create echo chambers, or push them towards propaganda.

But the upside is incredible. Integrating an LLM like GPT4 into a more kid-centered suite of models unlocks tons of potential.

The first thing I built into my AI tutor system was the ability for it to look up facts from reputable websites, and to help kids answer questions for themselves.

Then I taught the AI to help a kid develop their interests, so they take a topic they're curious about and dive deeper in to it.

And then I taught the AI to notify me if there are topics that are more appropriate for a parent to engage in.

The tool I'm building actually helps kids become more curious, creative, and thoughtful.

It isn't perfect yet, but the truth is kids spend 7 hours a day on screens already. I'm very excited at the possibilities.

Let me know if you'd like to try it!

seth@gotwonder.com

(I do realize this echos concerns from earlier generations about books, to video games and anything being bad for kids, but come on..)

I do think that we may be a ways off from truly being able to just "drop our kids off at the tutor," but I think we can see that, from here.

This energy... Would rather have Bing educating my children. I guess you've made your point.

I'm not saying this will _never_ be good. But it feels pretty early to unleash this bastion of confidently incorrect answers upon developing minds.

https://en.m.wikipedia.org/wiki/The_Veldt_(short_story)

I'll believe it when I see it.

(It makes sense -- if you assume the human connection isn't that important and people aren't social animals. /s)

Not even close, you sound like Libertarian-adjacent technologists who have been believing your own hype / some other equivilent metaphor.

Technology has a role to play in education that is as-yet unrealised and while I cannot speak for all of them, I know the system in place in the UK is behind the curve. The solution to this isn't to pull your kids out and use them as the cornerstone of an individualised experiment but to fight for better funding of the educational system itself. I hope this does not have any negative consequences for the kids who are the recipients of this trial.

My point is, we already ignore the obvious faults in education, science has been screaming about for many decades, for many decades. There must be some other reason why education is resistant to reform that would have to be adressed first.

But anyway, I would've really appreciated an AI as a kid that I could use to actually learn from after another useless school day with terrible, exhausted teachers, overcrowded classes, terrible textbooks and psychopathic class mates.

ME: what's trigonometry

GPTPlus: Trigonometry is a branch of mathematics that deals with the relationships between the angles and sides of triangles. It is used to solve problems involving triangles and is also applied in various fields such as physics, engineering, and astronomy. Trigonometric functions such as sine, cosine, tangent, cosecant, secant, and cotangent are used to calculate the values of angles and sides of triangles. These functions have important applications in fields such as navigation, architecture, and computer graphics. Trigonometry also has connections with complex numbers and calculus, making it a fundamental subject in mathematics.

Me: I need to use sine, cosine, and tangent. What are those. how do they work

GPTPlus: Sine, cosine, and tangent are the three most common trigonometric functions used in mathematics. They are abbreviated as sin, cos, and tan, respectively. Each of these functions relates the angle of a right triangle to the ratio of two of its sides.

Here's a brief explanation of each of these functions:

    Sine (sin): The sine of an angle in a right triangle is defined as the ratio of the length of the side opposite the angle to the length of the hypotenuse. In other words, sinθ = opposite/hypotenuse.

    Cosine (cos): The cosine of an angle in a right triangle is defined as the ratio of the length of the adjacent side to the length of the hypotenuse. In other words, cosθ = adjacent/hypotenuse.

    Tangent (tan): The tangent of an angle in a right triangle is defined as the ratio of the length of the side opposite the angle to the length of the adjacent side. In other words, tanθ = opposite/adjacent.

For example, if you know the length of the adjacent side and the angle, you can use the cosine function to find the length of the hypotenuse. Or, if you know the length of the opposite side and the adjacent side, you can use the tangent function to find the angle.

________________________

Me: can you go through an typical example of how to use sin with actual numbers

GPTPlus: Sure, I'd be happy to go through an example of how to use the sine function with actual numbers.

Let's say we have a right triangle with an angle of 30 degrees and a hypotenuse of 10 units. We want to find the length of the side opposite the 30 degree angle.

To use the sine function, we'll use the formula sinθ = opposite/hypotenuse, where θ is the angle of interest.

In this case, we have:

sin(30) = opposite/10

To solve for opposite, we can rearrange the formula:

opposite = sin(30) x 10

Now, we need to use a calculator or reference table to find the value of sin(30). The sine of 30 degrees is 0.5. So, we have:

opposite = 0.5 x 10

opposite = 5

Therefore, the length of the side opposite the 30 degree angle is 5 units.

This is just one example of how to use the sine function. Depending on the information given, you may need to use different trigonometric functions or formulas to solve for the unknown side or angle.

Yet, if the AI program can be copyleft then the matter is a non-issue.

I find this to be the heart of the contention.

I've built a tutor for my kids, that listens to them, and helps them explore topics that they're passionate about.

What's blown my mind is how it has driven creativity OFF screen too. They engage with the world in a deeper way when they have the power to explore their passions on their own.

If you want to try it, join the waitlist here: gotwonder.com and then email me: seth@gotwonder.com and I'll bump you up the list!