Live data from Hacker News

Building a real-time AI tutor for 5-year-olds

ello.com

211–220 of 460 posts

Re: Building a real-time AI tutor for 5-year-olds

#211
post #43
post #31

What are your thoughts about children in a Sudbury School model? These are democratic schools where children can do what they like in the day. Mostly they choose to play with other kids, games of imagination, though also doing screen time. One of the basic principles is that children figure out what they want to do and the learning comes along with it; the model views adults wanting children to learn something specif…

Hi! I'm Elizabeth (one of the co-founders of Ello). This is an interesting question. I actually think there is more overlap than people might assume, but it's a bit more because adaptability to various approaches to learning is part of the point. While I'm not deeply familiar with the Sudbury School model, I think there are various approaches to teaching kids that are successful because different approaches serve dif…

Thanks for replying. Glad to hear you use phonics. It sounds like it could have a lot of potential and I intend to pass it along to our community. If you would like to possibly explore its use in a Sudbury community, you can send an email to my gmail account which is the same username as listed here.

Re: Building a real-time AI tutor for 5-year-olds

#212
> A teacher is constantly deciding how to engage a student, whether to say something, draw on the whiteboard, play a game, or change topics entirely.

If it's really true that a teacher's job nowadays is to constantly try to cater to the whims of easily bored and distracted kids, then I pity both the teachers and the kids.

Re: Building a real-time AI tutor for 5-year-olds

#214
post #212

> A teacher is constantly deciding how to engage a student, whether to say something, draw on the whiteboard, play a game, or change topics entirely. If it's really true that a teacher's job nowadays is to constantly try to cater to the whims of easily bored and distracted kids, then I pity both the teachers and the kids.

This isn't a new thing though, nobody likes having to sit around all day and listen to someone talk and kids are the worst at that. They're young and have so much energy so a teacher has always needed to adjust the learning to make it something that the kids can learn from while paying attention and not disrupting the class. One of my favorite things my teacher once did after we read Gary Paulsen's Hatchet was have a 'survival expert' come in and talk about some of the things that Brian had to do and how some others (real people) had to do the same thing. The size comparison he showed us between a full grown man and a moose blew my mind as a kid.

Re: Building a real-time AI tutor for 5-year-olds

#215
The async planner is the part I'd love more detail on. If it's reasoning ahead while the kid is still talking, some chunk of that work gets invalidated every time the kid says something off-script, which at age 5 is presumably constantly. Is the speculative planning you throw away a real cost line or a rounding error? I build streaming LLM stuff (nothing this ambitious) and the "right move at the right moment" framing rings true, latency budgets change what the model can even attempt per turn.

Re: Building a real-time AI tutor for 5-year-olds

#216

I think teaching a child to trust an LLM from a formative age is horrifically irresponsible. If anything, an app should be made where a child learns to correct an LLM's mistakes and learn that it isn't trustworthy. Actually, better, don't put an LLM in front of children. At all. EDIT: If a use case is for children who can't afford good education, then use an LLM to make educational materials for children, review them…

I don't think it's quite as simple as that. LLMs do make mistakes but generally with slightly obscure difficult stuff. They're not going to struggle at all with things a 4-9 year old is learning.

Also it's not like teachers never make confident mistakes.

Re: Building a real-time AI tutor for 5-year-olds

#217
post #185

Earlier quoted context omitted.

Why?

Are you for real?

I'm not seeing the issue with e.g. Treasure Mountain or Treasure Mathstorm. They can do little math, reading, and logic puzzles while listening to Bach and Beethoven. It's engaging enough to get them to practice addition/subtraction but unengaging enough that they don't sit at it for more than 5-10 minutes at once.

Re: Building a real-time AI tutor for 5-year-olds

#218
post #121

Earlier quoted context omitted.

In practice, yes. In theory a machines can do about everything a human can, but better and cheaper. One of the main constraint for education is available tutor time, see e.g. Bloom's 2 sigma experiment. Obviously there are many pitfalls to overcome at the moment, but eventually machines will become better teachers than teachers, and not many parents will send their kids to public schools if the kids can learn much fa…

> In practice, yes. In theory a machines can do about everything a human can, but better and cheaper. If you could time-travel back to your 5y old self, would you prefer to be taught by AI tutor given the current state o/t art, or taught by whatever teachers you did have when you were 5? (with all the existing hallucination, breaking through guardrails etc problems of current AI in mind) If you'd have a ~5y old yours…

> If you could time-travel back to your 5y old self, would you prefer to be taught by AI tutor given the current state o/t art,

I don't know about Ello or whether is it better than human tutors yet.

> How do you see peer-to-peer contact in that scenario?

Neighbor kids gather and play as they please, which is also easier if they have more time on their hands, stay home, and overall live in each other's proximity.

Re: Building a real-time AI tutor for 5-year-olds

#219
post #31

What are your thoughts about children in a Sudbury School model? These are democratic schools where children can do what they like in the day. Mostly they choose to play with other kids, games of imagination, though also doing screen time. One of the basic principles is that children figure out what they want to do and the learning comes along with it; the model views adults wanting children to learn something specif…

Sounds a bit like unschooling which is a terrible idea because kids don't know what they need to know.

Think about what every adult absolutely needs to know: how to be an effective member of society, how to interact with people, how to self-regulate, and generally being a person other people want to be around and work with. That is what a Sudbury school supports learning while conventional school hinders it, offering an authoritarian-based lifestyle instead.

Most people going through conventional school don't get the freedom to learn to be themselves in a community until their 20s, not mastering it until often their early 30s. As for academics, Sudbury students have repeatedly demonstrated having no problem quickly learning whatever it is they want or need to learn as an adult. Many Sudbury students go to college, getting A's even in their first semester, making good bonds not only with other students but also faculty and administrators. They help form communities wherever they go. They are engaged in classes in a way that their peers are not. They are coming to academic learning fresh and eager not weighed down by 14,000 hours of conventional school tedium. Those who do not go to college often advance very quickly in whatever work they end up doing as person skills are the key to being successful.

The results speak for themselves. You can read about it in Peter Gray's book Free to Learn and other works by him.

Also keep in mind that teaching, which is what conventional school focuses on, is not the same as learning.

Re: Building a real-time AI tutor for 5-year-olds

#220

[flagged]

HN's skepticism on this subject (which I share) is understandable. At the same time I share your view of its enormous potential.

Personally I'd question the "who is this coming from?" (you answered that in another comment), and "what's the intention behind it?".

If you're a commercial company, I'd be tempted to suspect a "hook 'em while they're young" thought behind it. Perhaps not now, but turning into that over time because 'reasons' (cough "shareholder value" & the like).

If you were a non-profit otoh, I'd give you the benefit of the doubt.

To address that, would you be willing to pursue a non-profit status for your company and/or platform? Has this been discussed @ some point?

Btw: thanks for taking the time to address the many issues raised in this thread! HN can be a brutal place for people not used to its audience! (trust me, we're mostly friendly but with an extreme no-bullshit attitude at times).

Post reply on HN