AI for children and parents

Meetups
1

AI for Kids and Parents

We’d like to propose a project/topic with Irena which we both thought and talked about for a while. The question we are trying to answer is relatively simple - how do we make sure AI is safe to use for our children?

We might go and try to influence researchers and companies who build these LLM models, but the chance of that succeeding is slim. Where we might find a bit more success is education - explaining where and why LLMs fall apart (aka Factual correctness), why and how not to give your data to AI companies (aka Privacy Risks) and how to make sure we use AI in a way that it does not make us stupid (aka using AI for brainstorming, not brain-replacement).

We have come across many people asking similar questions - especially concerning children and students - and we believe we should be able to provide simple and clear answers and guidance on these topics.

Factual Correctness

We’ve all seen many examples where AI was simply wrong, yet 100% pushing the incorrect statement as a fact. We know why that is - LLMs are in essence just probabilistic bag of words generators wrapped in guidelines and rails to make them seem actually intelligent.

I like to say that AI is definitely artificial, but hardly intelligent.

We would like to find examples where AI relatively reproducibly fails (I’ve had plenty of cases where I give it a text and ask it to “make it nicer” and it completely changes meaning), explain the “temperature” setting which introduces randomness - great for your texts to be less boring, but also more…well…random:)

Generally to explain importance of fact checking when using AI and avoiding blind trust because “AI knows”.

Privacy

We could look at privacy policies of popular services and point out data that is being tracked. Highlight things like AI companies complying with police when they ask for your data etc.

Explain why giving all your inner thoughts to these big corporations is a risky action - especially when it comes to psychological issues, unconventional questions, drugs or any other matter where privacy is very important.

Another topic to cover would be integration of AI to other services - e.g. WhatsApp introduced AI integration which can freely scan your group chats unless you enable advanced privacy

Do not let AI make you stupid

Some research around this is popping up, but I believe many of us experienced it first hand - we are relying on AI for basic things which makes us think less. AI is amazing for brainstorming ideas, doing basic research into a new topic or reviewing what you have already thought about.

But many people start to offload their own thinking to LLMs - and it could potentially be dangerous especially for kids. It goes with the risk of not getting truthful information back, but also just not putting your brain through the “workout” and doing the hard things.

This is a sensitive one, so we need to find the right angle and wording:D

Distribution

We thought about what form should the output of this “exercise” take. It seems a website is most straightforward and versatile. As many of us here have kids, we can share it in our circles (pun intended) and communities and see if it is useful.

Then we could extract some basic points in a form of flyer that could be distributed to coffee shops, libraries, schools…

And last, we could also produce a workshop which we could offer to schools or parents.

Anyway, I am going to present this in Brno Circle and I believe Irena is planning the same for Berlin, but we are absolutely open to ideas and feedback - if you already know about materials covering these topics, maybe we do not need to do anything:D Or we just translate it to our own languages.

2 Likes
2

We brought this up at the Logos Circle Brno meetup yesterday and got 2 suggestions/comments:

  • It would be good to target children directly - i.e. focus not only on how parents should guide kids when using AI, but also directly speak to kids/teens - super valid point, I think it was in the back of our minds, but it is good to explicitly say it
  • Find influencer(s) with children audience and try to convince them to collaborate - provide them with content and let them do the explaining and distribution - I like this idea because simply building a website does not solve the problem of getting it in front of people, which working with some Youtube or something like that does.

We are very open to any thoughts, idea, comments, suggestions!

1 Like
3

You know… I had this very same conversation with a couple of good friends of mine who I studied with in University.

Couldn’t agree with you more though. The children are definitely the most vulnerable because they’re in the process of developing their critical thinking skills. If they have the option of choosing between having to suffer and persevere through understanding concepts or taking the path of least resistance, almost 100% of the time they won’t think twice about outsourcing the process of problem-solving to the LLM’s.

What’s worse is that there is usually no concern for understanding the process of how the answer was obtained. True comprehension of the material is set aside for finishing homework quickly. In the long run this is detrimental to their growth.

But honestly anyone and everyone is at risk of doing this. It all comes down to establishing good habits when interacting with AI/LLM’s. And not to be classist or elitist in any way, but the level of education or even how knowledgeable a person is, is a fairly good indicator of whether someone has the sophistication to properly use this tech.

This is great stuff and a very serious concern though :ok_hand:t4:
Let me know if I can help in any way.

Kindly,
Shaq

1 Like
4

How about, vote for politicians who do things like this:
From 10 December, children under the age of 16 will no longer be allowed to have social media accounts. The Australian government says the ban - a world-first policy popular with many parents - is aimed at protecting them from harmful content and other risks such as cyberbullying and grooming.

5

Sorry, I don’t follow - do propose we vote for politicians who will ban LLMs for children?

This is one thing I personally do not want - it is a role of parent to decide when their child is ready for any kind of technology and it should be upon their discretion - because they know their children, not the politicians.

Second, preventing people/kids using any technology or access any information can only lead to 2 outcomes -

  1. They’ll use whatever “illegal” way they can to access it
  2. If they don’t there is a significant potential they’ll lack essential skills for the future (in this particular case, using AI effectively)

I believe education - of both the parents and children - is the right path

1 Like
6

Not sure if he’s being sarcastic there with that response, but I’ll reply to @deca12x 's comment respectfully : ) !

I actually may have a pretty hot take here I feel… let people get their lumps out there, but prevent serious harm from occurring. Where I may deviate a bit from @vpavlin is that I strongly believe there is much value and good in children, teens, and adults to not be completely sheltered from the world. In a safe and controlled manner we have to build a healthy tolerance to the cruelness and horrors that come with life.

I do second @vpavlin point of view on NOT taking extreme courses of action like the one presented above. Those who’re intended to benefit from these policies are negatively impacted and are not helped. There is no effort to educate or intervene in a constructive manner.

Completely banning certain classes of people from social media will have them find other ways of gaining access to those platforms.

Education is a great foundational intervention. It doesn’t end there. More can be done :relieved:

It’s a very well intentioned decision; I don’t think it’s the correct one. Also, just to be clear, I’m not upset or responding to your reply in an aggressive way. Just conversing with you as I would a good friend.

Sincerely,
Shaq

7

Great topic that we must not overlook as our kids are going to grow with AI.

1 Like
8

I suppose that short form text like this is not ideal for these conversations - having circles is much better.
The example made is for social media - i.e. society’s first contact with AI. We have to account for contextual differences. LLMs are far more powerful and unpredictable. Their game theoretic Nash equilibrium is very harmful for society and especially for children. I don’t have the answer, but I want political representatives that are paying attention.
What I am most attracted by in the Australian Labour party is the courage and willingness to treat social media as a first tier issue. In the case of social media, it may be a more straightforward tradeoff, given the usecase is mainly addictive brainrot. For LLMs it will be more difficult, but even more important to have a public sector that treats the trend as front and center, due to AI porn, AI “therapy”, AI manipulation/nudging, AI slop, AI worsening ideological sylos, AI increasing human detachment…
Overall, Vaclav I see your point, I don’t think an underage ban is the best move in this case, but at the same time I hope for well researched political intervention attempts.

1 Like
9

Sounds like you’re headed the Cognitive Sovereignty/Cognitive Security route actually.

By any chance do you have any ideas or possible solutions for this problem ?

10

Yeah, absolutely, this is a tough topic and longer discussion is definitely needed:)

The problem with Australian law for me (one of them:)) is that they also ban Youtube - youtube is a great source of information and a way to learn. So can be social media - it just depends on how things are presented, how the algorithm works.

I 100% agree it is a mess and children need to be protected - my main point is that government is not the right body to do this.

I also understand not all children are well protected by their parents - not all parents are good parents and have time, energy or capacity to take care of this.

But I’ll reiterate - children are not stupid - I have spent hours upon hours chatting with random strangers on DC++ and elsewhere while waiting for all the pirated data to be downloaded… I used Tor to peak into the darknet when I was 14 or 15. Children and teenagers will do anything to try something that the adults say they should not do - you were a kid too:)

So from my point of view this is an inherently broken approach - it also teaches parents that they do not have to make tough choices and communicate them to the children - if I believe my kids should not have a phone, social media etc., I should be responsible for explaining that to them - not just offloading that decision and explanation to the government ban.

For me this is a deeper problem and by blindly banning things we are not treating the sickness, we are masking the symptoms.

If I bring it back to AI - fun problem there is that you can ban anything, but any kid with a gaming PC can easily run their own model and share it with his friends, so any bans there will have absolutely zero effect and the only viable solution is the education.

2 Likes
11

I too believe educating children/parents is the best path forward - this would allow people to understand how the tech works and decide for themselves whether it would work for them or not, and under what circumstances.

Perhaps it’d be a good idea to create content in Kurzgesagt (https://www.youtube.com/@kurzgesagt) style educating users on how LLMs and GenAI work. I and my classmates thoroughly enjoyed watching Kurzgesagt as children in school due to the animation style and narration. The educational nature of it is the icing on top.

The content could include how LLMs are trained, where data is taken from, good and bad examples of LLM outputs based on trained data.

12

Honestly, there is too much to say / respond to too many points made here (generally I agree with most things said here) - so I’ll defer to Tristan Harris - by far the person in AI safety that I most look up to. I highly recommend listening to what he has to say
I can’t post links here - search for this on youtube tristan+harris+center+for+humane+technology

13

I’ll check! Thanks for the pointer!

14

This is pretty cool and important as well. People ought to at-least know how the future might or will look like with over reliance on AI systems especially for kids, I also find myself in a similar boat once in a while when I’m feeling too lazy. In our own logos circle , we are also targeting education as well, but in our case it is more on the right penetration and sensitization… We brainstormed abit on the proper distribution channel/method as well.