Decoding AI
Tune in to this episode of Void Deck as AI expert Dr. Ayesha Khanna breaks down how AI works, its impact on jobs and education, and why seeing it as a tool that adds to human talent is key.
Episode 4: Decoding Artificial Intelligence with Dr. Ayesha Khanna (Addo AI)
In a world where AI is rapidly transforming industries, Singapore is positioning itself as a global AI hub. But what does this mean for our daily lives and jobs? We debug the complex code of artificial intelligence with Dr. Ayesha Khanna, co-founder and CEO of Addo AI and a globally recognized AI expert based in Singapore. Dr. Khanna shares insights on how AI is revolutionising businesses, the impact of AI in the classroom, why seeing AI as adding to and not replacing human talent is crucial in the future of work. Plus, we hear whether sci-fi films like The Terminator showing a robot uprising are onto something.
Listen to the Full Episode
Available on: Spotify, Apple Podcasts , and Amazon Music.
Subscribe to Void Deck wherever you get your podcasts to get notified when the next episode drops.
Episode Highlights
Decipher how generative AI works - from Shakespeare to FIFA, discover how generative AI mixes and matches vast amounts of data to create something new
Explore the concept of "AI friends" and why they might be both wonderful and dangerous - it's not just about chatbots anymore!
Hear whether there’s any truth to popular science fiction movies–will AI take over the world?
Be warned about "AI hallucinations" and why you shouldn't trust everything ChatGPT tells you (especially in court!)
Get the scoop on why naming your robot something cute might be the key to embracing how AI can be for you
Timestamps
00:00 Preview
00:39 Intro
01:40 Dr. Khanna’s educational background
03:35 How Dr. Khanna got into AI
07:41 Dr. Khanna's work at Addo AI
10:13 AI applications in healthcare, insurance, customer service
13:46 What jobs are at risk of being replaced by AI?
16:55 [Mid-roll Ad] Smart Nation Playscape at Science Centre Singapore
17:29 Is the next generation at a disadvantage if they become reliant on AI?
20:42 The dangers and wonders of AI friends
23:21 Paiseh question: will robots/AI take over the world?
26:35 Paiseh question: how does generative AI really work?
30:07 Word Association Game
34:35 Outro
Guest Biography
Dr. Ayesha Khanna, co-founder and CEO of Addo, is a leading AI expert and Forbes-recognized entrepreneur with 20+ years of experience in digital transformation. Holding degrees from Harvard, Columbia, and the London School of Economics, and having spent over a decade on Wall Street, she advises CEOs on AI and data transformations and serves on boards such as Johnson Controls, NEOM Tonomus, and L'Oreal's scientific advisory board. A passionate advocate for inclusivity and diversity, Dr. Khanna provides AI education scholarships to girls and mid-career women through her education company, Amplify.
Transcript
This transcript has been lightly edited for readability.
RISHII: Will AI take over the world?
DR. KHANNA: I think that AI on its own will not get to that point because by that time humans will have merged with AI in some way.
RISHII & MENG: Whoa.
DR. KHANNA: So, you know, and we're seeing something like this with Neuralink, obviously, what Elon Musk started. So the question is, are there humans who will supercharge themselves with more information, more cognitive power, and will that create a superclass that will then be at such an advantage that the others are left behind?
RISHII: Welcome to Void Deck, a casual science podcast brought to you by Science Centre Singapore. We bring science out of the labs and put scientists in singlets. Each episode, we sit down with a local science changemaker and ask all the questions that you are too paiseh to ask. Today, I'm joined by my colleague Meng, a science educator at Science Centre Singapore.
MENG: Hi everyone.
RISHII: Hi, and today we'll be talking all about AI, something you and I probably use very often like, you know, ChatGPT and all those cool stuff.
MENG: Today on the podcast, we have Dr. Ayesha Khanna, CEO of Addo AI and a Forbes-recognised AI expert. She has over 20 years of experience working on digital transformation with degrees from Harvard, Columbia and London School of Economics. Dr. Aisha advises CEOs and serves on major boards. Through her company, Amplify, she offers AI scholarships to girls and women.
DR. KHANNA: Hi everyone, it's such a pleasure to be here.
RISHII: Thank you so much for coming down today, you know, to talk to us about AI and all the cool stuff that you do.
DR. KHANNA: Thank you for having me.
RISHII: All right, so I think what we want to do is take you back down memory lane and see how you got started on this journey. What was it like growing up, you know, in Pakistan and like, you know, getting into the field of STEM as a kid?
DR. KHANNA: Actually, I was never in the field of STEM. When I was growing up, I was much more interested in literature and human rights and development economics. I found science to be taught in a terrible way. It was all about the exam and about the absolute correct answer. Even when we went to do quote unquote experiments in the lab, we already knew what the experiment should yield.
So that's a terrible way to learn science. It's so deterministic. And it was only when I went to college, when I left Asia, that I went and encountered people from Romania, Estonia, Russia, and they had a very different approach to science. They looked at it in a very poetic way.
They saw it not as the final answer in an engineering mechanistic way, but they found engineering, technology, science, beautiful, uncertain, ambivalent, and the whole process as a quest to approximate the truth. And when you see it that way, how different is STEM from the humanities? How different is it from sociology or any of the social sciences?
And the answer is it isn't. Which is why anyone at whatever stage in their life, if they just realise this about STEM, can easily step into STEM and any STEM person can easily step into any other kind of discipline.
RISHII: I just wanted to know, because I mean, you made the jump from like a humanities person and then you went into STEM in college, right? When did you even consider AI as something that you wanted to pursue? Because AI [seems] like something that probably started maybe five to ten years ago. So at that time when you were in college, did you already [know] about AI or was it something you just stumbled upon?
DR. KHANNA: Well, AI has been around for decades.
MENG: We were wrong. Very wrong. [Everyone laughs]
DR. KHANNA: It's been around for decades, but nobody took it seriously because we just did not have the computational power that we do now. We didn't have enough data that we do now. And so it was kind of languishing at MIT and other kinds of universities. I first encountered neural networks decades ago myself as an undergraduate when I was working on neural networks for trading.
And then when I went to Wall Street, I was doing a lot of quantitative work where we were using computer science and statistics. Today, the combination of computer science and statistics at scale is called artificial intelligence, which is, can you recognise patterns in something and then be able to very quickly mine large amounts of data to be able to analyse it, forecast it, optimise it, and now for the first time generate from scratch new things. We call that generative AI.
But it's not new, but it is an evolving field. And it goes through summers and winters or rather springs and winters. So there was a whole time when it was called the winter of AI and now it's having a real moment. But who knows, after 10 years, it might go through another winter where people lose interest or feel it's not moving fast enough.
RISHII: Right, right.
DR. KHANNA: The thing about science and AI is science, right? It's a scientific endeavour… it takes time. Then people work and work and work and suddenly something great happens and it's adopted widely. And then again, you have to wait for it for a long time. So there have been AI researchers diligently working on it and I've been exposed to it over many, many years. Now I'm so happy to see that everyone is realising it and finds it more accessible.
RISHII: It's like, you know, if you can't like me on my low days, don’t like me when I'm super popular now. So when you were interested in AI, right, like when you stumbled upon it in college and you're telling people like, oh my God, AI is going to be big or I'm so interested in it, were you like the only one who seemed to be interested, maybe in your friend group or in your social circles?
MENG: Or was it like a big crowd and everyone was going to jump onto AI together?
DR. KHANNA: No, no. At that time, people didn't even understand what it was. I told people I was going to become a software engineer. And there were people in my extended family, they said, oh my God, we sent you to Harvard and you should be in a bank, like a posh office. And now you're acting like a back office person. And they were so like pooh-poohing it.
RISHII: All the uncle and aunties.
DR. KHANNA: All the uncle and aunties thought that I had really taken a step down because they didn't understand it. And it's only now that it's suddenly become in the media that they—the same uncles who used to criticise my parents and say, why did you let her do this?—now actually send me articles and are very excited about AI. And I guess the moral of the story is, do pursue what you really want to and work at it.
So one doesn't go without the other. You can't just be passionate about something and then just do it half-heartedly. I stuck to it through many, many years, even though it was hard and nobody really saw it. But I loved it because every day I could see the impact it had on the firms that I was working with.
[Musical transition]
RISHII: Do you get, I guess now when you're working on AI, do people come up to you and think that you are working on like ChatGPT or just like generative AI? Do people still just make the assumption? Because I think our exposure to AI, or at least with the layman, right, is just like ChatGPT and that's all we know.
MENG: Generative art… deepfakes…
DR. KHANNA: That’s right.
MENG: Stuff like that.
DR. KHANNA: Well, people know that I have an AI company, I have an AI consulting firm. We specialise in building AI strategy, AI solutions, big data platforms. And so we use generative AI to help our clients engage their customers better, for example. But they never come to me because I'm not a researcher, even though I have a PhD and have a lot of experience. I am not a researcher.
My job is to help organisations, whether they are for-profit companies, non-for-profit enterprises or government agencies, think about how to use AI to serve their customers better. And so that's what people ask me about. They're like, how can we use this? What's the talent we need? Is it true that it is a hype or is it real? Those are the kind of questions. Are our jobs going to get replaced? Are we going to have a job?
Should my son study this? Should my daughter study this? And I feel these are great questions and we should talk about it. You know, is this ethical or not? These are the debates citizens must have, regardless of whether they have a science background, a business background, any background.
RISHII: Wow, that's very well put. On that note, we really want to know more about what you do in terms of AI. We understand that your company is an AI solutions firm and it's called Addo. So how did that name come about?
DR. KHANNA: It's very simple. It's adding. So Ado is what we wanted to say that we augment, catalyse, enrich and accelerate the journey of our company with AI. So we are really there. And that's something I really believe in.
RISHII: Ah, okay.
DR. KHANNA: We need to have that approach, not only at a corporate level, but also as an individual. We need to see AI as adding, not replacing us in any way. And so just by that name, that's our mission, is to use it as an augmentation tool for all our customers.
RISHII: You were talking about helping your customers. Would there be any unusual or surprising industry applications that you have come across in your time working with all these different partners and companies that seek your help?
DR. KHANNA: I think there's so many examples, but just let's look at a couple of them. One example is that there are so many people in the world who die of or suffer from strokes. It's terrible. I mean, when anyone has a stroke and it's so unexpected. People suddenly one day and then they lose their ability to walk, they lose their ability to talk. And it's devastating for the family.
It's devastating for everyone around them and for themselves. So Mayo Clinic actually partnered recently with an AI startup that analysed the medical records of all the patients who've had stroke so that they could identify a pattern. And in that pattern, they found that they could actually identify someone who will have a stroke 10 years before they're going to have a stroke.
Can you imagine if people knew that? If you know that, wouldn't you change everything? Because these are modifiable biomarkers. So this is the kind of thing, the kind of massive game changer that AI can do to help people live better lives. And there are many examples like this, but I find health care to be the most potent example because we can all relate to it.
RISHII: Meng and I were talking [about] how else AI helps people. Because I think on a day to day basis, we use generative AI for simpler tasks like LinkedIn captions.
MENG: Finding my way to work. That's also AI, right? I'm sure, we're using GPS and other stuff.
DR. KHANNA: Oh, absolutely. We use AI all the time. But most people don't understand or appreciate that generative AI is fundamentally new in the AI field. We have never been able to generate new content from scratch like this. That's why people are so excited. And that means, for example, I'll give you [something] that we worked on, insurance companies.
One of the things that people call insurance companies for are pregnant women who've just found out they have preeclampsia and they're really worried for their child. And at the other end is a customer service agent who's probably young, doesn't have children, is rifling through all the PDFs or something's being told to this person that, OK, say this about the policy. They want to check if their policy covers this health condition.
But now if you have an AI agent, it can generate with the voice, multimodal, which means not just text, but a voice. It can give a very reassuring, empathetic answer that can surpass that potentially of the contact centre agent. Personalise it, be more empathetic and listen to the emotions in that person's voice because then you can tell, is this person worried? Is this person upset? Is she angry?
And this is called hyper-personalization real time. So as long as they keep the data of the person safe and private, we will be able to enter a new world of hyper-personalization where every company seems really interested and catered to our needs. And I think that is going to change customer experience drastically and improve it.
MENG: But when you say that…how should I put it? There's always that aspect of, is this going to replace the people? And you've mentioned it yourself. What jobs would be, for the lack of a better phrase, at risk of being replaced?
DR. KHANNA: Many, many tasks will be at risk of being replaced. And I think that's the right way to think about it. So we can't put our head in the sand like an ostrich and pretend like nothing will get replaced. But we also can't think big, like jobs will get replaced. Actually, there will be certain things that all of us do.
And I'm sure if we thought back at our day or yesterday, there are lots of things that we do that an AI could do for us, like scheduling meetings or drafting emails or LinkedIn, as you were saying. LinkedIn postings. And this is not just about productivity. I often write articles this way as well, but then I improve them with my own voice. It makes the whole process more fun, more interesting.
So some of the time that we spent on certain tasks, they don't only have to be routine, they can be creative as well, we’ll have less time needed, because generative AI will help us. And it's not just about time, but it's also about that whole process. Kind of us having like a co-pilot, like an assistant, like a friend. So we'd actually be able to have a more interesting experience also.
The problem is companies don't know what to do with that 20 percent time that people have free now. So there are two things to do. One is just freak out and say, oh, my God, I'm just going to fire people. That is the wrong approach. The right approach is to take that 20 percent time that we'll all have and show companies that we can do something else with it. We can add value. We can analyse the firm. We can analyse our customers. We can think about strategy.
And leaders should never let go of employees who understand their market, their customers, because what a loss. You have to train somebody from scratch again. You should use those people for more strategic tasks. And that's what we're seeing. The great companies, they use AI to actually grow and then they keep their people.
But what happens, I think the danger is junior people, because you don't know we need so much junior staff anymore.
So now we need to then you're like, oh, so are my kids going to do or you guys are still young? What are you going to do?
And so I think that at that point we have to change the education system so that when these graduates like my kids come and look for a job, they already are so well versed in working with AI that now they're saying we can do X plus Y because we already know how to work 20 percent faster and we'll do something else as well.
So it's that repositioning of one's value proposition as an employee that I think is very important. And I don't think anybody who knows how to work with AI will lose their job in the near future.
[Mid-roll ad starts]
JAMIE (Void Deck Producer): Want to continue your TechWizard journey? Visit the Smart Nation playscape at Science Centre Singapore. You'll find a digital playground with eight zones of fun, including the AI zone, where you'll teach a computer how to recognise your doodles and 3D animals. You can even grow your own digital plant to discover the exciting world of augmented reality. For more awe-inspiring visuals, step into E3, an immersive exhibition featuring virtual reality and 360-degree projection mapping. Get your tickets at science.edu.sg. Now, back to the show.
[Mid-roll ad ends]
RISHII: We're seeing the next crop of generations who probably use AI almost on a day to day basis. Do you think they're at a disadvantage because now they're using AI? I mean, they might become reliant on AI. Let's say I'm a copywriter. I learned the fundamentals of copywriting and now AI is kind of like making that job easier for me. But if, let's say, I was in school still and I'm using AI more to copywrite, would I now lack that fundamental skills of learning it in the first place?
DR. KHANNA: I think so. That's true. I think the education system has to teach people to be answerable for how they critically evaluate, think and creatively construct. It could be stories, experiments, science. And that means that the AI that they're allowed to use is really there as an assistant and not giving them answers.
So, for example, if you look at Salman Khan's Khan Academy, which is really well known in the United States and provides free courses to kids, he has a ChatGPT version that he calls Khan Migo. And Khan Migo will not give children the answer. In fact, it will be like a Socrates, like it will have a debate with them. So they will say, we want the answer to this for my homework. And I'll say, no, no, no. Did you try this? And they'll be like, yeah, we tried that. It didn't work.
So it is that engagement. It is more an Oxford/Cambridge-style tutorial system and less that Asian exam system. It's more about really understanding thoughtfully. But the difference is we'll almost never be alone. We'll always have a copilot with us. And so we'll get used to it. But we don't want to let go of our own critical thinking because humans have this ability to leap forward, make leaps of perception and imagination.
And that's wonderful. It gives us meaning. So I think that we will need to introduce those kinds of co-pilots in our education system also. And when we do, it would be fantastic for our kids.
MENG: So it's more like learning how to use a tool, right? You won't be doing the actual–you definitely will still be doing the actual product, but you need to learn how to use the tool to help you with the actual product. Is that [right]? So you're learning the skills of using the tool.
DR. KHANNA: You're learning the skills of using the tool, yes, but it's not like you're not learning. Yes, yes. You're not learning about the actual final end product. Yes. So you're learning the skills, for example, if you had to evaluate art, for example. So you're not just learning how to ask ChatGPT about it, but you're learning how do you aesthetically think about a painting. And then ChatGPT is helping you or questioning you or provoking you or inciting you.
And how great is that? Because who gets that one on one attention from a teacher? And the other thing is, which you said, which is very interesting, is that, yes, it's a tool, but it's kind of more than a tool now.
MENG: It's becoming part of everyday life.
RISHII: It's almost integral to a lot of things.
DR. KHANNA: Or more than that, like a friend. And that's both like dangerous and wonderful.
MENG: Yes, yes, yes.
DR. KHANNA: So if you think about character AI that was just bought by Google or many of these chat bots, or even if you look at older adults, there are millions of boyfriends and girlfriends that people rely on. And there was one article in The Washington Post where the woman said she preferred her virtual boyfriend to her real boyfriend. Oh, no. Because her real boyfriend was full of himself. Her virtual boyfriend was coded to be always listening.
But more than that, it's if I have three human colleagues and one AI colleague, I'll start to take that AI colleague seriously. Maybe I'll begin to have, you know, emotions or feelings for that person. And Sherry Turkle from MIT said that as humans, we kind of can't help ourselves. But if anything's animate or displays emotions, we kind of become very susceptible to liking it.
And so the question is, I fundamentally don't think there's anything wrong with having AI friends. But we must govern the people who produce these AI friends, because if they want to, they can manipulate us.
So if, you know, I'm going to have a boyfriend and suddenly he says, well, why don't you go and, you know, pick Max Factor blush? And then I do. Or why don't you take a loan to get this new house? And I do. That's an influence that is beyond just helping me. It is now a commercial interest. So these are the dangers, which is why governance is so important.
RISHII: Right. There's so many ethical boundaries, right?
DR. AYESHA & MENG: Yes!
RISHII: Even with that example that you gave with the virtual boyfriend and the actual boyfriend, I think there's some moral dilemma for them as well. Like, oh, is it like, are you cheating on me?
MENG: I'm losing my job to an AI!
DR. KHANNA: Oh, that's also true, are you cheating on me? How interesting. Very true.
RISHII: Do you see it as an actual human person, or do you see it as an AI? Then suddenly there's a whole new layer of looking at it.
DR. KHANNA: And if you look at like, if you look at chat GPT 4.0 and you know, the audio version of it or Gemini Live, literally you cannot tell if it's a human or it's a bot. It's funny. You can interrupt it. This was a big problem before. You can interrupt it. Like I could just change the subject and both of you would go with it. Right. But usually the AI would freak out because it's like going like a robot. Now the AI will be like, oh, okay. Yeah. And change.
[Musical transition]
RISHII: I think that would be the perfect segue to our paiseh question segment where basically every episode we ask a paiseh question, something we might be a little embarrassed to ask to somebody working in STEM. My first paiseh question is, will AI take over the world? Because I used to be–not I used to be–I love the Terminator franchise.
And the whole concept was about AI becoming self-aware and taking over the defence system and slowly the whole world. Do you think that is accurate? Will we see some semblance of it in the future? Or do you think it's just, you know, just sci-fi, it's never going to happen?
DR. KHANNA: No, I think that AI on its own will not get to that point because by that time humans will have merged with AI in some way. So, you know, and we're seeing something like this with Neuralink, obviously what Elon Musk started. So the question is, are there humans who will supercharge themselves with more information, more cognitive power?
And will that create a super class that will then be at such an advantage that the others are left behind? And the only way to stop this is for governments to come in and regulate the sector. We're nowhere near this right now. Maybe our children, our grandchildren will face this.
But certainly if AI continues to get very intelligent and it can connect to the brain of certain individuals and now you have, you know, you have exoskeletons, people can do so much. So, I think the bigger question is, will a class of superhumans emerge?
RISHII & MENG: Whoa.
DR. KHANNA: Which will be for their own interest. And a lot of movies are about this issue, right?
RISHII & MENG: Yes, yes.
DR. KHANNA: Because if you ask me, the AI is not immoral or I mean, it's humans, right? Who kind of make it that way or train it that way. So right now it's not conscious enough. But I think at that point we really need to think about these things and have these circuit breakers. Where we don't let it get to that point.
I expect the European Union will be the first to think about this when there's any danger of it because they're fearless in regulating AI, even if it means stifling a little bit of innovation. But Singapore has a very balanced approach. We're very pragmatic. We give guidelines. We are not free for all. Sometimes like some Western countries, nor are we very restrictive. And I think we always have to walk that fine line.
And the more people participate in this, like, you know, there's a youth policy forum that Prime Minister Lawrence Wong has put together. The more we talk about this as a country together, the more we can make sure that we stay on that line where we benefit from all the innovations that come from AI, but really keep it away from its manipulative or maleficent actors that may control it.
[Musical transition]
MENG: I have a question.
RISHII: Is it a paiseh question?
MENG: It is a paiseh question. Because I myself, I don't really understand this very well. We know what generative AI is, but how does it generate? How does it work? How can you explain it to, say, the layman?
RISHII: Because it really feels like someone's just sitting on the other end, waiting to answer all my questions.
DR. KHANNA: Well, look, I think the easiest way to understand it is that if you look at, let's just do text. And first of all, it's important to know that generative AI is not only for text. It can generate everything. So eventually it can generate, it already does sound, images, video, text, and then it can do new kinds of proteins, new kinds of antibiotics, new kind of chemicals. It's doing a lot of things.
But essentially what it's doing is taking a lot of the information, let's say, as we know, with ChatGPT, it went out, it got all of the text in the world on the internet. And now if you ask it a question, let's say about Shakespeare, or you say, write me something about, you know, FIFA, football, and the way Shakespeare would write it. So now it needs to put together a lot of things.
It needs to put together, know all of Shakespeare's plays, plus kind of know how Shakespeare would have reacted to sports, know all the sports that's happening. And it's this ability to take the important patterns, pay attention to the important patterns that it sees in all this text, and very, very high speed, generate something out of it. So it's taking everything it knows. It's only picking out the important things.
That's why the big paper that set it off was called “Attention is All You Need,” because it pays attention in the story to the important things that it needs to remember, and then generate something for you. It's the same thing if you gave it a bazillion pictures of a cat, then it can eventually know if the picture is of a cat or not, and also make the picture of a cat.
But now if you give it, say, give me a picture of a cat sipping a martini, then it has a lot of other pictures of people sipping martinis, and so it can combine the two together to inform it. But the interesting thing now is the big question, as I was telling you earlier, is does it actually form some concepts in its mind of how a person sits, so the physics of something? I don't know if that's true, but that's what they're finding.
Give it enough data, like human beings, like children, we begin to form some concepts. But again, it's not conscious because children can do so much more. The big problem with generative AI is that it needs so much data foundationally, or has in the past anyway.
MENG: It's moving so incredibly fast.
DR. KHANNA: It is moving fast, and I think that it'll only get faster. But nobody needs to worry about it. The worst thing is to feel overwhelmed by the speed because there's so much noise. Fundamentally, if one had understanding that these tools serve you and your company and your career and your children, it's a lot about mindset, believe it or not.
If everybody listening right now has a mindset that this is for me, your whole attitude changes.
RISHII & MENG: That's true.
DR. KHANNA: But it requires some willpower. It's kind of like going to the gym. [Everyone laughs] Yeah. You know, you're like, I should go, I should go, I should learn AI, I should work with that tool.
RISHII: Maybe tomorrow. [Everyone laughs]
DR. KHANNA: Yeah, maybe tomorrow. But please do it today.
[Musical transition plays]
RISHII: We want to play a short game with you. It's a word association game. So pretty much what happens is we'll be throwing out some words at you. And you need to tell us what comes to mind. OK, so I'm just going to go with one that's really weird. Robocop. [Laughs]
DR. KHANNA: Dystopia.
RISHII: Do you think we will have actual robocops in the future? You know, you were talking about supercharged humans.
DR. KHANNA: We will have robotic policemen or robo soldiers, but robocop has had a very negative connotation. Right. That's the reason why I say dystopia. But I have a friend who was just visiting and his robot is actually being used in ports and fire departments all over the world and is doing security rounds and helping those departments. Is it a robodog? Yes. But because we have such a bad association with it, we shouldn't use those words.
We should try to use other words. Maybe call it a security dog or something. That's what the Japanese do. They always have cute names for everything. And I asked someone when I went to Japan, I said, you know, we have terrible associations with robots. You know, and he said, yeah, that's why we make cute things so people are not intimidated and have cute names for it. And to be honest, it made a lot of sense because then you don't feel intimidated. And in fact, you feel you have agency and you're empowered.
And when you're like that, you can work with it and you can control it. As opposed to being afraid of it.
RISHII: You're right. Perception thing.
DR. KHANNA: Mindset.
MENG: Took a while to get my mom used to the little robot vacuum cleaner. So I gave it a name and added a little googly eyes on it.
RISHII: What do you call it?
MENG: It's Bob. His name is Bob.
RISHII: Harmless.
MENG: My turn. So this is a word that's quite poignant to me. We have a lab named after this [person]. Ada Lovelace.
DR. KHANNA: Oh, my daughter. My daughter dressed up as Ada Lovelace in her, I think it was grade three, because she loves AI and Ada Lovelace is a big hero of hers. And she dressed up because she thought that it was such a good role model. And I, whenever I think about Ada Lovelace, I think of her.
MENG: Oh, that's so cute. By the way, our computer lab is called Lovelace Lab.
DR. KHANNA: Oh, how nice. Perfect.
MENG: Yeah, I hang around there too much. But I love that place.
RISHII: My next word would be “hallucination.”
DR. KHANNA: Generative AI.
RISHII: Oh!
DR. KHANNA: When you ask a chatbot a question, it can make mistakes. And the word that is used in the industry is that it's “hallucinating.”
RISHII & MENG: Ohh.
DR. KHANNA: And a lot of people work on reducing the hallucinations of generative AI. So the last thing you want is that someone is talking to an AI and it's giving made up answers and they're very good processes and systems.
And we use that such as RAG [retrieval augmented generation], etc. to control it. So these are called risk guardrails. But people who are not careful, there was this, can really get swayed by hallucinations. So there was a lawyer who was lazy, I suppose, and covered in the New York Times. And he was late with preparing his defence. So he asked generative AI, GPT, and he came up with all these, you know, you have to give precedence. Yes. He said, my Lord, you know, in this case, this happened. And in this case, this happened.
And the judge was like, dude, like, this is all fake. Oh, no, none of this ever happened. You get out of my court right now. Oh, my God. So I always tend to double check, spot check what AI tells me. And that's a critical thinking hat that I have on. But over time, we will have an AI check the work of another AI and then we won't have to worry about that.
RISHII: It's like AI-ception now.
DR. KHANNA: Yeah. I mean, now we have agents, you know, checking other agents work, but we're not there yet. But reducing hallucinations is very important because people tend to trust what they think is some entity smarter than them.
Whereas AI is not smarter than humans because it is not like humans. It's machine intelligence. And we have human intelligence. And it's very important for us to see it differently. I mean, I have a dog who I adore. I don't think he has like, I don't think he acts like me. He has dog intelligence. We really need to separate these things a bit so that we remain in our own responsible, creative thinking, critical thinking beings.
[Musical transition]
RISHII: So thank you, Dr. Ayesha, for joining us. I think it's been so great, not just like learning about your interest in AI, but and, you know, like all the misconceptions people have about AI and all the cool stuff that might happen, like supercharged humans. I really hope that comes to precedence and I will be one of the first.
DR. KHANNA: Thank you so much. It's been such a pleasure. I'm so honoured and grateful to be on this podcast. I'm a huge fan of the Science Centre.
RISHII: Yes!
MENG: Please come to the Science Centre.
DR. KHANNA: So, carry on. I will be there. Thank you so much.
RISHII: So thank you so much. It was such an exciting episode. And if this episode has sparked your curiosity about future technologies, visit Science Centre Singapore to explore the cutting edge world of artificial intelligence. Discover digital playground with eight zones of fun at the Smart Nation Playscape Exhibition. Afterwards, watch robots in action at the Object Theatre in the Future Makers Exhibition. All of these await you at Science Centre Singapore.
MENG: Follow Void Deck on Spotify, Apple Podcasts and wherever you get your podcasts. To find out more about Dr. Ayesha’s work, you can follow her on LinkedIn and explore her website at ayeshakhanna.com.
RISHII: See you next episode.
MENG: See you next episode.
Resources
Dr. Ayesha Khanna's Website
Khanmigo by Khan Academy
Mayo Clinic research finds AI-enabled ECGs may identify patients at greater risk of stroke, cognitive decline
China’s lonely hearts reboot online romance with artificial intelligence
https://www.washingtonpost.com/world/2021/08/06/china-online-dating-love-replika/
Vaswani, Ashish; Shazeer, Noam; Parmar, Niki; Uszkoreit, Jakob; Jones, Llion; Gomez, Aidan N; Kaiser, Łukasz; Polosukhin, Illia (2017). "Attention is All you Need". Advances in Neural Information Processing Systems. 30. Curran Associates, Inc.
https://arxiv.org/abs/1706.03762
This Artificial Intelligence Learns like a Baby
MIT sociologist Sherry Turkle on the psychological impacts of bot relationships
Here’s What Happens When Your Lawyer Uses ChatGPT
https://www.nytimes.com/2023/05/27/nyregion/avianca-airline-lawsuit-chatgpt.html
Want to continue your AI learning IRL? Check out our tech-related exhibitions at Science Centre Singapore:
Future Makers
https://www.science.edu.sg/whats-on/exhibitions/future-makers
Smart Nation PlayScape
https://www.science.edu.sg/whats-on/exhibitions/smart-nation-playscape
E3: E-mmersive Experiential Environments
https://www.science.edu.sg/whats-on/exhibitions/e3
Credits
This episode of Void Deck was hosted by Rishii Vijayahkumar and Meng Hwee Lim. The episode was produced, written, edited, and sound engineered by Jamie Uy. Sound recording and post-production assistance was provided by Lydia Konig and Joyce Sia. The episode graphics were designed by Vanessa Ng and Jamie Uy. The podcast cover art was illustrated by Vikki Li Qi. The background music "Data Flow" and "Spatial" was created by Fugu Vibes. Special thanks to Dr. Khanna for coming on the show.
Last updated 17 December 2024