Tshilidzi Marwala | Use AI responsibly for a better world (S13-E7)

 

[00:00:00] Melissa Fleming

My guest this week grew up in a small village in apartheid South Africa and went on to specialize in artificial intelligence.

 

[00:00:09] Tshilidzi Marwala

I would not trust AI. You are not supposed to trust AI. You are supposed to use it. Always doubt it. Always verify what it tells you. Cross-validate so that you can be able to use it responsibly.

 

[00:00:39] Melissa Fleming

Tshilidzi Marwala is rector of the United Nations University. From the United Nations, I'm Melissa Fleming. This is Awake at Night. Welcome. It's really great to be doing this interview with you here at United Nations University headquarters.

 

[00:01:07] Tshilidzi Marwala

Thank you very much, Melissa and thank you very much for coming to Tokyo. Tokyo is a fascinating city. Of course, one of the biggest cities in the world. And we are glad you are here.

 

[00:01:19] Melissa Fleming

You have had the most amazing life. You are from South Africa, where I understand you grew up in a small rural village. Can you describe where you were born.

 

[00:01:31] Tshilidzi Marwala

I was born in the part of South Africa called Venda. And Venda is the little part of South Africa that is north of the Tropic of Capricorn. So South Africa is maybe 95% south of the Tropic of Capricorn. So I was born in that part of the world. I grew up in a village called Duthuni. I love going there. I love the energy. I love the people.

 

[00:02:03] Melissa Fleming

What did it look like?

 

 

[00:02:04] Tshilidzi Marwala 

Well, you know, I always say that I lived in a period of transition because Venda was almost an isolated place. Of all the places in South Africa, it was the most isolated, and for a number of reasons. The main reason was that it was a malaria area. And as a result, we would be the group of people in South Africa to actually last see Westerners. Of course, the people who live on the coast, in the Cape, Xhosaland or Zululand. You are talking about... It can even go as far as 600 years ago. Vasco da Gama, when he was trying to go to the east, came through there in the 1490s.

So I grew up there. Interesting because just north of it if you cross the river, you are in Zimbabwe. And of course, it was during the time of apartheid. And of course one of the principles of apartheid was separate development. So Venda was declared a republic.

 

[00:03:23] Melissa Fleming

Were there any white people?

 

[00:03:25] Tshilidzi Marwala

Of course there will be. I would even say that when I grew up, the ratio probably was coming down. And the reason for that was because when separate development came, the ones who were in Venda, they had to leave Venda because of this new policy, you know. Almost to the point that my father converted to Christianity and had me baptized when I was still very, very young by a German Lutheran priest, Pastor Landman. And I can sort of remember him, and then he just disappeared. So that was where I grew up. And of course, like any rural part of South Africa, you have a chief. And our chief was a very strong education advocate. He was a great chief.

 

[00:04:33] Melissa Fleming

What did your family do?

 

[00:04:35] Tshilidzi Marwala

Well, my father was a math teacher who was absolutely a very, very crucial, pillar of our community. The first generation in his family to actually go to school, which was very, very good. And my mother took care of us.

 

[00:04:58] Melissa Fleming

So you had a chief who really cared about education. Your dad was a teacher. I believe you had a big favourite in your family and that was your grandmother.

 

[00:05:14] Tshilidzi Marwala

Yes. You see, my grandmother did quite a number of things for me. The first one, I think she made me grow up much earlier than I grew up. And I always make this joke whenever I go back. So, when I was young, in the morning, a neighbour would come almost every day and come and start gossiping about what is happening in the village. I don't know whether they thought I could not understand. You know, and it would be about disputes in a home. It would be about fights. It would... Whatever, weddings that are coming, you know, and so on and so forth. I would hear that. So that's the first thing that she actually taught me.

 

[00:06:05] Melissa Fleming

So the things that adults probably would want to have hidden from curious children.

 

[00:06:13] Tshilidzi Marwala

Yes. I heard it. And I can even remember her voice, this neighbour who used to come. The first thing that she starts with is, 'Have you heard?' You know, and in my language, you say [inaudible], you know. When she arrives: 'Have you heard?' And then she will start, and there will be a long story and so on and so forth about maybe disputes in a family and so... So, I was young. And so I understood complexities of human interactions and conflict at quite a young age. And the second thing, she really taught me about how to connect with nature. She used to make clay pots. And I would accompany her to the river to collect clay. Clay is almost always next to the river. She'll come back. She'll form this very beautiful pot.

 

[00:07:11] Melissa Fleming

By hand.

 

[00:07:13] Tshilidzi Marwala

By hand. And then she will put it in the sun so that it can dry. She will put it in the furnace so that it is absolutely strong. And after that she will knock each pot and she will listen to it very, very carefully.

 

[00:07:30] Melissa Fleming

With her finger she would knock on it?

 

[00:07:32] Tshilidzi Marwala 

And then she will say, 'This is a good pot. This is a bad pot. This is good pot if it rings for a long time. It's a good pot.' And I go to Cleveland, Ohio as an undergraduate mechanical engineering student. I take a course called dynamics. And in dynamics, you learn about vibration. And the teacher who taught me was Professor Joe Prahl, an old American gentleman. And then, of course, he teaches us about damping. Same thing. If you knock an object that rings for a long time, it has a low damping coefficient. If it rings for a short time, it has a high damping coefficient. And of course now that I understand engineering, the reason why the one that rings for a short time is not a good pot is because there might be things trapped inside. It might be even fluid. It might even be things that are not supposed to be in the pot.

So I really appreciated that she was able to connect with nature. She was a very important pillar in my life. Unfortunately, she passed on when I was doing first year in Cleveland. So I did not get to attend the funeral. Telegram arrived. Telegrams were... They were an efficient way of communication, but they were bad communicators. The reason why they were bad communicators is because you had to pay for every word. So it will say... It was abbreviated. The message just said, 'Tshianeo is dead. Stop.' And I miss her every day.

 

 

[00:09:31] Melissa Fleming

Did she inspire you to become a scientist?

 

[00:09:34] Tshilidzi Marwala 

Yes. You know, because later on, when I arrived at the University of Cambridge as a PhD student, I did not really know what I'm going to do a PhD on. I was really open-minded. I had to actually go and see, of all the things that I have studied, what am I interested in? Obviously, it was not going to be acting. So it was engineering.

So, I arrived there. So my PhD was on fault identification using neural networks. That is the title of my PhD thesis. Of course, neural networks is the dominant version of AI. So basically, you have all these structures and systems that are always communicating something. You just need to know how to identify what they're saying. It can be a network, a telecommunication network, and just the presence of a cyber-attack changes the dynamics of the network, and you can observe it, and you be able to shut down things so that you can protect the people.

Same concept, you know, in many critical structures. They just put things on the surface of the structure, and they monitor and now with AI the system will be able to determine whether the system is good or not. Whether it is a human body in hospitals, you have all these measuring devices that are going to some AI system, and it will actually be able to tell you whether the patient is critical or not.

So that's actually what I did my PhD on. Same concept as my grandmother. Instead of using my grandmother's ear to listen to the sound, you just put an actuator, a measuring device that measures what is happening. Instead of my grandmother's brain, which later I found out that it was not perfect because as time went on, she started throwing away good pots. And the reason was because her hearing was actually deteriorating because of age related reasons, you know.

So I went and did the same for my PhD. So you would have whatever instrument you're going to be monitoring. You take it to some computer. And then I built an AI system that looks at the signal that is coming and classifies what it is seeing. Is this structure healthy or not? That is the first classification. If the answer is yes, good. If the answer is no, can you classify what type of a fault this system is actually experiencing? And then, of course, a machine is able to do this at a much, much smaller scale than a human ear. And an AI system is able to understand not just one signal, but a thousand signals because all you need is just computational power. So that is a direct link to my grandmother and ultimately to my PhD thesis.

 

[00:13:28] Melissa Fleming

So this is quite a journey from being a small rural village boy with no technology to getting your PhD in AI. I mean, just taking you back, you must have studied really hard. Because I understand that when you were 18 you won South Africa's National Youth Science Olympiad.

 

[00:13:55] Tshilidzi Marwala

Yes.

 

[00:13:56] Melissa Fleming

So how did you even know to apply to that and what did that mean to you?

 

[00:14:03] Tshilidzi Marwala

I think when I arrived in high school, there was a crisis in our education system. Because especially in urban areas, African kids were out of the classrooms, on the streets. And then of course, that was the time when the then president of South Africa said he is going to make a big speech. And then, of course he was going to... You know, there's this expression about Julius Caesar that he crossed the Rubicon. It was going to be called the Rubicon speech. And then of course, it never really happened. The president, the then president came and said, 'We are not going to be pushed.' And then there was a big crisis. So things started changing, especially for people like me, who live next to the borders. So we had to be watched, obviously, you know.

 

[00:15:18] Melissa Fleming

So you're saying there were lots of protests then. What year was this?

 

[00:15:25] Tshilidzi Marwala

'85, '86. I mean, '85, I think half of the African kids did not go to school because in urban areas schools were completely stopped. And a lot of people came to the villages. Villages are always slow. They are slow. Things don't come.

 

[00:15:42] Melissa Fleming

But it didn't also... Apartheid didn't reach them as dramatically.

 

 

[00:15:47] Tshilidzi Marwala

Yes, of course. Then we started having teachers who came from places that we did not know. From India. From the Philippines. And also conscripted young Afrikaans boys and English boys coming to become our teachers.

 

[00:16:16] Melissa Fleming

Why was that?

 

[00:16:17] Tshilidzi Marwala 

Well, I mean, the only way I can explain it is that we lived in the border, so we needed to be... You know, borders had to be secured. And if you have opposing forces, you know, the borders are a bit of a risk area. But also, it was clear that times have changed and things need to change. So all of a sudden, we had all these teachers. And some were very, very good, I must confess.

So the school, one of the things that was introduced was that we will start writing the Science Olympiad. So I wrote the Science Olympiad. And then as I was told that I was going to go and that as a result, I'm going to attend a science conference in Johannesburg. I was really, really happy. It was my first time going to Johannesburg. I'm really a rural boy. It was the first time going to Johannesburg. They said, 'You're going to be in Johannesburg for the week. And then you are going to go to the United Kingdom for two weeks.'

 

[00:17:21] Melissa Fleming

You must have been excited.

 

[00:17:22] Tshilidzi Marwala

Really excited, you know. And then, of course, I had to go and apply for the passport. And I went to the United Kingdom. And there were a number of things that were defining. The first thing that was defining was that at that conference there were delegates from East Germany and West Germany representing two different countries. And this must have been like three weeks before the fall of the Berlin Wall.

 

[00:17:58] Melissa Fleming

So this was in 1989.

 

[00:18:01] Tshilidzi Marwala

Yes. Literally, when the Berlin Wall fell, maybe three weeks back, I was in the room with people from two different countries - East Germany and West Germany. Interesting, you know. Now, I often wonder, did they know what was going on? Did they know that the wall was going to come down? And then, of course... And then the conference was done at a college in South Kensington called Imperial College. Little did I know that that will be the first place I will start my academic career after my PhD. One of the sessions that we had at that conference was a session on what are the technologies that are going to change the future. It was not AI.

 

[00:19:01] Melissa Fleming

No. What was it?

 

[00:19:02] Tshilidzi Marwala

It was superconductivity.

 

[00:19:06] Melissa Fleming

Okay, well, we're going to get into that because I know what you'll say is the technology of the future and the present because you went on to study it. But you ended up going to the US first, didn't you?

 

[00:19:21] Tshilidzi Marwala

Yeah, my first degree was in the US. When I was in England, I quite liked the international vibe of London. I really liked... I loved it. I loved the energy. I loved the freedom. And the first time I arrived at Trafalgar Square, because there was the South African Embassy. We went to the South African Embassy. And at that time, there used to be 24-hour protests outside the South African Embassy. The freedom was just something that I had never actually been exposed to.

 

[00:19:57] Melissa Fleming

The freedom to express oneself politically.

 

[00:20:00] Tshilidzi Marwala

Well, I think it is... Of course, that is a very important part of freedom. But there is another form of freedom, and I don't want to be controversial. If I go home, there are certain elements that I feel I am not free. There's a certain level of expectation on how you talk and what you do and so on and so forth, you know. And sometimes people, especially people who live in liberal democracies, don't understand how much sometimes just norms restrict you quite a great deal, you know. But they are just certain things that I could never say to my father.

 

[00:20:55] Melissa Fleming

You mean community norms. There were certain rules and etiquette.

 

[00:20:57] Tshilidzi Marwala

Yeah, there's just something, you know. I mean, whenever my father was saying something that I knew was not correct my biggest problem was, how am I going to frame it in such a way that he does not think I'm saying he's wrong, but he knows he's wrong? Of course, that gives you the ability to be creative, you know.

 

[00:21:34] Melissa Fleming

I think that would be a useful method anywhere.

 

[00:21:38] Tshilidzi Marwala

Exactly. So you know, I mean, the first thing was when young people would stand up to the lecturers and say, 'I don't think you are right.' It's a very important component of productivity.

 

[00:21:55] Melissa Fleming

I mean, I studied in Ohio. It was very close to your university, Case Western. I went to Oberlin. It was so cold in the winter. I mean coming from South Africa, how did you...? That must have been a shock. You were young. First time in the US, but in the middle of Cleveland.

 

[00:22:18] Tshilidzi Marwala

Yes, absolutely. You see, I had never really seen snow. South Africa has snow, of course. Lesotho, you can even ski, because it is sometimes 3,000 metres above sea level. You go to the Drakensberg. It snows. Johannesburg snows once every 10 years. One day, you know. Not heavy snow, you know. But at our place it does not snow. But we have a word for snow. So my people have seen snow somewhere. Otherwise they would not invent… Snow is called mahada. So they would not have invented the word. So there is snow. But when I arrived at Cleveland, it was when you just wake up in the morning and there is snow. Bit of a shock, you know. And then, of course, what I really appreciated was the multidisciplinary educational experience.

 

[00:23:32] Melissa Fleming

You were away from South Africa during a very critical point in history. By the time you graduated, the apartheid system had been dismantled.

 

[00:23:43] Tshilidzi Marwala

Yes.

 

[00:23:44] Melissa Fleming

What did that feel like for you?

 

[00:23:52] Tshilidzi Marwala

You know, when I went back... I was in the United States for three years without going back to South Africa. And there were elections in 1994. And me and a friend of mine from Oregon, a gentleman called Brian, said, 'Let's go and do a tour in Southern Africa.' So we arrived in South Africa. It was not too long after the election. The election was... Freedom Day is 27 April. So May, after May we were in South Africa. So it must have been like three weeks after we arrived there, you know.

 

[00:24:35] Melissa Fleming

But it must have been exhilarating. And what did you think of Nelson Mandela?

 

[00:24:39] Tshilidzi Marwala

Yeah, it was. Absolutely. And by the way, I never really met Nelson Mandela directly. Never had a one-on-one with him. In fact, I can’t even say I met [inaudible]. So it was very, very nice. And then we arrived there, and then we realized that freedom is beautiful, but it's a very difficult journey. Democracy is a beautiful concept, but it's a very difficult journey. It's not an overnight thing.

 

[00:25:16] Melissa Fleming

Back to your evolution. You ended up leaving South Africa and going back to the UK to do a PhD in Cambridge. I was just wondering, you had just come from freed South Africa, and then you were going to Britain. Did you have any feelings or apprehensions about being in a former colonial power?

 

[00:25:42] Tshilidzi Marwala

Look, Britain is very fascinating for South Africans. Absolutely very fascinating because the bonds between Britain and South Africa are long. We really are... Have long bonds, you know. Winston Churchill was a prisoner in South Africa during the Anglo-Boer War. He even wrote a book about it, you know. So going to Britain is of course going to a foreign land but there is something very familiar about it. It was fascinating going to United States. But going to England, sometimes we used to go there so that we could understand better who we are. To understand our flags. To understand our political systems. All these issues, you understand them better when you go to the UK.

 

 

[00:27:07] Melissa Fleming

Interesting. What's really interesting is that you chose to do your PhD in artificial intelligence long, long before it was widely talked about. I'm curious what made you choose that.

 

[00:27:20] Tshilidzi Marwala

No, I mean, I actually came to AI by accident. So when I arrived at Cambridge, I was not really quite sure what I wanted to study. And I said, 'What am I going to study?' And then at one stage I was sitting during lunch, sitting next to somebody from South Africa called Nando de Freitas. He is a Vice President now of artificial intelligence at Microsoft. And then I say... From Johannesburg. And I say, 'Nando, what are you studying?' He said he's studying artificial intelligence. I'm like, 'Oh, what is that?' And then he explained to me. 'Oh, that is what I'm going to study.'

 

[00:28:07] Melissa Fleming

You weren't intimidated by it at that stage because this was really early on. I mean, nobody was talking about AI.

 

[00:28:15] Tshilidzi Marwala

Yes. I mean, AI is actually quite old. The term was invented in 1956 at Dartmouth by John McCarthy. Alan Turing had already written his paper in 1950 - "Can machines think?" Of course, he was coming from the Enigma machine and so on and so forth, you know. And most of the algorithms... I always say that there is not a single algorithm that I can see, maybe except generative adversarial networks, as a structure. But it still uses neural networks, two neural networks that are competing with one another. There aren’t not too many architectures that I see today that I did not see in the 80s. But the two things that were limited - was the amount of data that was available and the amount of computational power that was available on the data.

 

[00:29:30] Melissa Fleming

What did you tell your parents what you were studying and how did you explain it?

 

[00:29:37] Tshilidzi Marwala

My aunt says to me, 'So what are you studying?' I tried to explain it. And she just couldn't make sense of what I was saying. She says, 'I think you don't understand what you're studying.' I said, 'Why?' She says, 'But you can't explain it the way I can be able to understand it.' It was very difficult to explain.

 

[00:29:56] Melissa Fleming

But they must have been proud of you when you came back to work at the University of Johannesburg where you became... You were professor, then vice chancellor, then principal. And then you introduced an artificial intelligence course there for students.

 

[00:30:14] Tshilidzi Marwala

Absolutely. You know, I mean, I'm proud of my students.

 

[00:30:18] Melissa Fleming

What promise does it have for education? I know you're really hopeful about harnessing AI for development. And I know also when a lot of people speak about, 'Okay, so what good can AI possibly do?' They cite, actually, education.

 

[00:30:36] Tshilidzi Marwala

Absolutely. I think the idea of making education accessible in every way possible is an important issue.

 

[00:30:47] Melissa Fleming

So somebody is in a village, and they have very little access to education. How can AI help?

 

 

[00:30:52] Tshilidzi Marwala

Well, maybe more than just AI. Because I have my own skepticism, you know. I think databases, even just Wikipedia, have been a very valuable source of information.

 

[00:31:14] Melissa Fleming

Which has not been run by AI.

 

[00:31:18] Tshilidzi Marwala

No, no, no.

 

[00:31:19] Melissa Fleming

In fact, AI threatens it.

 

[00:31:22] Tshilidzi Marwala

No, no, no. AI... The training of large language models actually relies heavily on Wikipedia. If you were to go and ask AI, 'Tell me about Adam Smith.' I mean, there must be lots of Adam Smiths. But there is one Adam Smith that has a huge Wikipedia page, the one who wrote "The Wealth of Nations." So AI works very good for things that are mainstream.

 

[00:31:59] Melissa Fleming

On the other hand, there are those who know this very well, and they're creating content that they know that AI will feed on that is not accurate.

 

[00:32:14] Tshilidzi Marwala

Yes. That in itself is not an issue, because we always know that AI is not accurate.

 

[00:32:26] Melissa Fleming

But ordinary people don't.

 

[00:32:29] Tshilidzi Marwala

Yeah, I mean, when you study...

 

[00:32:30] Melissa Fleming

They think it's all knowing and they trust it.

 

[00:32:32] Tshilidzi Marwala

No, no, no. When you study a course in AI, they have a theorem. It's called PAC theorem - probably approximately correct model. I'm not making this up. You can go and check it. This is a theory that you are told because AI relies on data and data is never certain. It's never complete by definition, you know.

 

[00:33:05] Melissa Fleming

But it also relies on the internet, and the internet is full of garbage.

 

[00:33:11] Tshilidzi Marwala

No, of course.

 

[00:33:12] Melissa Fleming

The Nobel Prize-winning economist Joseph Stiglitz calls it garbage in garbage out.

 

[00:33:19] Tshilidzi Marwala

No, absolutely. So we know that it's not a high-precision technology. That we know. And as a result, we should always verify. The idea of verifying what it tells us is extremely important. That is why AI literacy course is very important. That is why here at the United Nations University we have created an online free course that is on LinkedIn learning.

 

[00:34:00] Melissa Fleming

Just quickly, what would you tell somebody who's skeptical about AI for good? And you really care about development, you said. What can AI do to supercharge development or the Sustainable Development Goals?

 

[00:34:21] Tshilidzi Marwala

No, I think it must be used in agriculture.

 

[00:34:27] Melissa Fleming

How?

 

[00:34:29] Tshilidzi Marwala 

Like for example, how to do soil analysis and already we have technologies like that. In places where labs are much more expensive to do, you can do a drone that takes a picture and be able to give you the soil analysis, and farmers can be able to use for crop rotation for example. Not as good as if you take the sample and go and analyze it in the lab, but it's not that bad, you know.

In hospitals, common diseases are going to be detected. Very, very important. Breast cancer is a big deal, especially in our country. And part of the reason is because they don't get to do mammograms. It's very expensive. There are techniques that can be able to detect, not as good as a mammogram, just electric conductivity and taken to AI and it is able to obviously assess the risk. Because it's not perfect, it will send a lot of people to a doctor. Well, at least the worst it can do is to send somebody you are not supposed to send to a doctor to a doctor, but it will be able to screen and be able tell you. It's cheaper making health accessible to people. Very, very important. If you look at issues of food production, Zero Hunger is a very noble goal. I think AI can play a big role.

 

 

[00:36:34] Melissa Fleming

When you think about AI, what is keeping you awake at night?

 

[00:36:42] Tshilidzi Marwala 

I mean, there are lots of things. I think the first thing that keeps me awake is how much the people who are governing AI are missing the crucial things. And I'm going just talk about maybe four of them.

The first one is what I term governance arbitrage. That you have this system that is supposed to be governed - AI. It has to be governed at the international level, and UN is doing that. But it's not only UN, it's also the International Standards Organization that has to do the standardizations. And then you have governance that has to be done at Member States levels. And you have AI that has be governed at the industry level. So you have all this and they are not exactly connected. So you have arbitrage where people who do not like a regulation in country A will simply move to country B and make a regulation there. I think that problem is actually ignored.

The second one - the algorithmic limitations and their impact on AI governance. I have already talked about rare events. We know that AI uses maximum likelihood approach, which basically means by design, it will marginalize things that are not highly represented.

And then the third thing is what I call rational opacity, the fact that we are supposed to trust systems that are sometimes operating so fast that we cannot be able to even understand what they are saying. And people are actually trusting these systems. What do we do about that? What needs to be done about that? And I don't think we are looking at it.

And then the last thing that I think is quite important is that we view AI as this single thing but it's actually four things. And for us to be able to govern it well, we should understand that it is four things. The first is that it's actually about data. It's about data. Secondly, it's about algorithms. And then the third thing is the computing. We tend to think of GPUs because it's globally... That is what the global competition is about - NVIDIA chips. But it's much more than that. Issues around investing in technology that make these tools energy efficient. It's not even on the table because the payoff is very good. Even if it is expensive companies will still go and make a lot of money, so they don't worry about how expensive it is. And then fourthly, how you apply it. And I think being able to always frame the governance processes around these four issues actually will be able to make us govern it better than we are when we're just looking at it as one singular organism.

 

[00:40:38] Melissa Fleming

You once said that technology is easier to fix than people. What did you mean by that?

 

[00:40:44] Tshilidzi Marwala 

Oh, wow. Yes. No, I mean, people are complicated. If I were to describe them in mathematical terms, they are not even non-linear. Sometimes they can be chaotic. Chaos, because a person can know what needs to be done and still chooses to do the direct opposite.

Whereas an AI system that is designed well and you design it to know what needs to be done, it will do what needs to be done. The problem is not whether it will do what it is supposed to do. It is whether it will have the means to know what needs to be done. And we can fix it. And once you fix a technology, it will be reliable. You go and put it in the... Once you train it, it is well-trained and you put it in production, it's going to produce.

Whereas a human being can be the most trained and knowing the job and then halfway through, they will sneak around and go and call a friend and so on and so forth and try to take shortcuts. Even lose attention because they're thinking about what is going to happen after work and so and so forth. That's really what I meant.

 

[00:42:24] Melissa Fleming

You're dealing with a lot of human beings here in Japan because you are the rector of the United Nations University. Can you just explain what this university does?

 

[00:42:35] Tshilidzi Marwala

Well, the United Nations University is supposed to do two things. The first one is to be a think tank for the United Nations System and the Member States. And the second one is to be an educational institution that trains people. That is really what is codified. But when I came to this rule, I realized that there is a third mission that is also just as important, which is to be an avenue of taking UN values to the people.

 

[00:43:21] Melissa Fleming

You are making sense of a lot yourself. Do you have children?

 

[00:43:25] Tshilidzi Marwala

Yes, I have three children. I have one girl who is 13 years old. She goes to school in Tokyo. I have a boy who is 15 years old, who is also here in Tokyo. And we have a son who is 23 years old, who is a graduate student in computer science at Rochester Institute of Technology in New York.

 

[00:44:02] Melissa Fleming

I figured that at least one of them was going in your scientific direction.

 

[00:44:07] Tshilidzi Marwala

And he is obviously now preparing... He wants to do a PhD. He's preparing to do a PhD now. We always... In my family, we do have quite a lot of engineers who have studied to the highest levels. But it's the first time that I have somebody that I can actually talk [to] about algorithms.

 

[00:44:42] Melissa Fleming

How great! And it's your son. Is this what you do to relax, or are there any other things you do?

 

[00:44:47] Tshilidzi Marwala

No, I walk a lot in Tokyo. I go to the gym a lot now. And then I go to opera here in Tokyo, beautiful opera. I love music and musical plays are some of the things that I cherish the most.

 

 

[00:45:21] Melissa Fleming

Lovely. Thank you, Tshilidzi.

 

[00:45:21] Tshilidzi Marwala

Thank you very much.

 

[00:45:22] Melissa Fleming

I really enjoyed our conversation. I feel like I learned so much. I feel more optimistic about AI too.

 

[00:45:29] Tshilidzi Marwala

Well, I mean, we should always be, you know... I would not trust AI. You're not supposed to trust AI. You are supposed to use it. Always doubt it. Because you always doubt it, always verify what it tells you. Cross-validate so that you can be able to use it responsibly.

 

[00:46:01] Melissa Fleming 

Thank you for listening to Awake at Night. We'll be back soon with more incredible and inspiring stories from people working against huge challenges to make this world a better and safer place.

To find out more about the series and the extraordinary people featured, do visit un.org/awake-at-night. Subscribe wherever you get your podcasts and please take the time to review us. It helps more people to find the show.

Thanks to my editor Bethany Bell and to my colleagues at the UN: Katerina Kitidi, Roberta Politi, Geneva Damayanti, Abby Vardeleon, Alison Corbet, Laura Rodriguez de Castro, Eric Balgley, Jason Candler, Benji Candelario, Anzhelika Devis, Tulin Battikhi, Bissera Kostova, Brianna Rowe, Joon Park, and Taeyoung Lee. The original music for this podcast was written and performed by Nadine Shah and produced by Ben Hillier.