Amanpour and Company

Well, as we have been discussing from deepfake images to voice scams, the danger of artificial intelligence has prompted calls for regulation.

Our next guest a season upside it it says it could be the biggest positive transformation education has ever seen.

He is the founder of the nonprofit Kan Academy, that recently piloted a tutor and a teaching assistant power to buy AI.

He tells us how he thinks it could supercharge world-class education.

>> welcome back to the show.

>> thank you.

>> we have heard about the promise and peril of artificial intelligence, but one of the amazing things that it seems to be doing, and you are leading the way on this, is creating a personal tutor so that every kid on the planet can have a personal tutor to tutor them in math, writing, history and everything else.

Explain how you are trying to do that.

>> educators have known for millennia that one-on-one tutoring, working with a student at their own pace, is the best way to learn.

That is what Alexander the great had with Aristotle.

Fast forward, we had a utopian idea of mass education, but we had to compromise.

We did not have the resources for personal tutor's, so we put students together and would have somebody lecture.

That is what we have been doing.

Over the last many decades, there has been tons of efficacy research that it is great to have 30 kids in a classroom but it would be better to have one-on-one tutoring.

If you do that commute could take the average student and make them an exceptional student.

Take a below average student and make them an above average student.

Many folks in technology over the last several decades have thought about how to use technology to emulate what a tutor would do.

That is what we have been doing for the last 15 years at the Academy.

Our not-for-profit mission is free education everywhere.

When OpenAI reached out to us last summer, we are under an NDA until only a few weeks ago, and they said, we want to do positive use cases.

We said, we think it is ready to actually hit the holy grail of education, which is can we create -- can we create a tutor for every child?

We launched in March.

And what we started piloting is our artificially intelligent tutor powered by AI.

And there has been news about using ChatGPT.

This does not allow you to cheat.

If you ask a question, it will say, how would you approach it?

It asked like Eric -- ask like Aristotle -- acts like Aristotle would.

It works across every subject.

Has all the context of the student would normally have.

And it does act as a teaching assistant for teachers.

>> give me an example.

Take history.

Suppose it is in American history course.

You and I have done some together.

How would it help us figure out how the Constitution was written?

>> we did user testing with many students.

We have a lab school, we have an online school, and we did research with these students.

And one of the students was looking at a part of AP U.S. government on judicial review, or judicial confirmation.

She asked the AI.

She watched a video and asked, why it is this relevant to right now?

And it immediately brought up recent confirmation hearings, Merrick Garland, that situation and all that.

The student said, wow, this just brought it to life in ways I could have never imagined.

It also allows students to do things that would have looked like science-fiction a year ago, where they can actually talk to simulations of historic characters.

You can debate federalism with Madison or with Hamilton.

>> school boards in the country are being rattled by all kinds of controversies about what is getting taught, about inclusivity or diversity or history or reparations or sexuality, gender issues.

How do you think a tutor, an AI tutor, would deal with such controversies?

>> you know, I might be nalve, but I feel like 95% of Americans are reasonably consistent on these issues.

And many of the polarizations happen on hearsay.

I heard of that is not in the classroom.

It gets folks triggered and angry.

One of our strategies has always been to just be hyper transparent.

If somebody tells you there is a video that is biased at the Academy, or are artificial intelligence is, show us and to show somebody else.

It is a secret.

It's tehere for anybody to try.

We will add just if we feel like there is a bias.

Or if it is trying to give a point of view that may not be fair.

I know folks at OpenAI, Google and other places are trying hard to make the underlying engines unbiased.

We have had actually some of the content you and I have done together, there is an appellate court judges that is skeptical of the Academy coming out of California.

He looked at our product on the Constitution and it said this is the way it should be taught.

And had another civil rights leader who said, have you papered over certain aspects of history?

He looked at the content and said, no, this is a full treatment on American history.

I believe when both sides can see it in its totality and did they say it is academic, not biased, I think most folks get behind it.

>> let's talk about math.

When somebody gets a math problem wrong, there are 100 different ways they can have it wrong.

Let's say there is a piece of algebra where you do not understand the distributive process.

How does a one on one tutor help you in that way?

>> as many folks know, the AI coming out, these are language models.

People have been skeptical about how good they will be at mathematics.

Even when we saw some of the first examples with GPT, it was doing well in the humanities, well in the science and conceptual knowledge, but not so well with math.

We have been working closely into spending time internally, also with researchers at OpenAI, to try to get the math right.

So, people are surprised by how good it is at math.

It is not perfect.

But if a student -- if they ask how to do the problem, it will not tell you, it will ask you what is the next step.

If the student does not distribute a property correctly, it will -- we have a concept called AI thoughts, where it will think about how it would have approached the problem or how the student could have a personal problem.

It does not share that but compares the student's response to that.

At that is what a tutor would do , think about it.

If the student did something different than what the AI thinks is a reasonable path, the AI will say I have something different.

Can you explain your reasoning?

That is a very good pedagogical thing to do.

When the student explains, they can understand it better.

So this is what a great tutor would do.

I have something different than you.

Instead of saying this is how you do it, say explain your reasoning.

Then 80% of the time the student will say, this is where I messed up.

Or the AI might say, that was actually a better approach.

>> for it to be transformative it has got to be equitable.

It cannot help increase the divide between rich and poor.

So how do you think about that?

What extent do you hope and invasion it might be free for every kid on the planet?

>> this is the core issue.

One of the major sources of an equity -- an equity in education is you could have two students that go to the same classroom, but one student whose parents understand the system, who have access to resources, they might get real tutoring when they go home.

Or their parents could tutor them while the other student does not have access to resources.

The idea of having a scalable tutor is to level the playing field.

So, we think already, even with the computation costs, it is far more accessible than traditional tutoring.

And then, if the cost per student keeps going the way it looks like it will, we think in the coming years it will truly be something we can give to every student and every teacher as a teaching assistant.

>> you talk about personalization.

It fascinates me because when Aristotle was tutoring Alexander the great, it was personalized.

He knew what Alexander was having trouble with, what he had difficulties with a year ago.

To what extent will this thing remember you throughout your entire school career, and be personalized for you?

>> that is what we are literally working on as we speak.

If you were to use it right now, it remembers the conversation and it remembers some of the more traditional work you have been doing.

By back to school, we hope it will remember its conversations it has had with you.

It will remember if you told it I prefer this kind of tone, or it knows your reading level.

So, this is not some science fiction even three or five areas out, it is more like three months out that that will be there.

And we will continue to make it more and more personalized.

And we will be running ethics studies.

We have done studies on the core of the Academy, but now we will see adding a layer of artificial intelligence can really accelerate students, not just learning, but likely their engagement.

>> that sounds awesome, but there is one possible DarkSide to it remembering everything about you and being totally personalized, which is privacy.

Do you have a guardrail or guard rails so that I could not subpoena or nobody will be able to get the private data it has?

>> yeah, that is core to who we are.

15 years ago, when I set it up as a not-for-profit, and I did not envision that AI would advance this quickly, but one reason we were not is be recognized student data is a sensitive thing.

And we wanted, amongst other things, our true North to never use data for anything that could be counterproductive.

It should only be used to improve learning for the student, personalize it, or improve the efficacy of the platform.

These are things we have taken seriously.

Even with the current AI, we are not using that information to train artificial intelligence.

Some of the questions, I think for students there is a different context.

This is one of the safety mechanisms, that everything a student does is monitored by the teacher and parent.We have another artificial intelligence monitoring conversations with the first artificial intelligence to flag conversations and notify parents and teachers.

To your point, we definitely come over the coming years, especially as it gets knowledge of the student, make sure that it is only used for positive use cases.

>> we recently had Jeffrey Hinton on the show and he said that there could be sort of an existential threat of AI.

What's your perception of that?

>> I think it is hard for us to predict exactly what will happen, but the thought experiment I run in my mind is there is the more conservative stance on AI, where we say we do not know where it is going, let's regulate it before we actually see it cause problems.

The problem with that is the only people who will follow that are the good folks, the rule followers.

The criminal organizations, the authoritarian states, they will not slow down.

In my mind, the most dystopian scenario is one where criminal organizations have better artificial intelligence than the rule followers.

I do not think it is a viable path to act with fear and try to slow things down.

The other path, folks will say this will work out, like the Industrial Revolution.

It will create more jobs than it destroys.

May be, but I do not think it is enough to hope.

And use it as a strategy.

I think it is important for actors like our nonprofit and others to be very proactive and to say, what are the risks?

How do we mitigate them?

What are the benefits and how do we maximize them so AI becomes a massive positive as opposed to a negative?

>> how do you envision 10 years from now education?

>> I think you will be able to talk to a tutor, an artificial intelligent tutor, like we are having a conversation right now.

It might even happen in five years.

It will be able to draw things out, like a real time personalized video.

It will happen in any language.

And you might sometimes engage with it on your phone, laptop, or through virtual reality so you feel like you are in the same room with it.

It will be immersive.

I thank you also see changes to other parts of -- think you will also see changes to the other parts of the system.

How you can grade, where you have multiple-choice, and that is what the educational system got focused on, things you could assess in a low-cost way.

But now artificial intelligence can't assess riding, thoughts can you can have a simulation or dialogue with it.

You can have an oral exam with it, which is the gold standard for a thesis defense.

You can do that on demand.

In five or 10 years, assessment will be looking a lot richer.

I think that the teacher's role in this artificially intelligent world, a lot of administrative tasks will be taken away and they will be able to focus on the one to one personal attention.

They will always have the artificial intelligence to help advise.

We have another nonprofit called schoolhouse.world and we are already leveraging artificial intelligence to give the tutor feedback on how they can tutor better based on the transcript.

And we are about one year away from in real time being able to give the tutor feedback on, hey, you did not call on this student lately.

Here is what they are actually asking.

So it will really be something where it is not humans versus artificial intelligence.

It will be artificial intelligence allowing the humans to be more human.

>> should colleges, when they decide to admit students, should they be allowed to submit all of that so a college can say, this is how a person works.

What about job applications?

Is that in invasion of privacy or is it something that would be useful so college admissions would be more fair?

>> anything is reasonable as long as the people who are affected by it bought into it.

I could imagine a world where a student interacts with conmigo over many years, maybe their entire k-12 experience.

Then they could ask conmigo to write a recommendation for them.

It is the student asking for it.

I would not want to do it behind the scenes.

I would not be surprised if University admissions, going back to your question, if they have to sift through 30,000 applications.

They have readers.

It will be inconsistent depending.

But they will be.

using this kind of technology As well as people know how they are using it, testing it for bias.

As long as it is more perfect than what we are doing today it is a step in the right direction.

>> thank you for joining us.