AI Interrogator: Navigate Business AI with Shamala Sadananda

Kate Bevan: Hello and welcome to this episode of the Infosys Knowledge Institute's podcast on AI, the AI Interrogator. I'm Kate Bevan of the Infosys Knowledge Institute. My guest today is Shamala Sadananda, who works with me at Infosys. She is a tech champion. She's been bringing innovation and emerging techs to our clients for 20 years and specifically working with AI. Shamala, thank you so much for joining us. It's a real pleasure to have you.

Shamala Sadananda: Kate, thank you for having me. It's always such a pleasure to talk to you.

Kate Bevan: You have such an overview of the whole tech scene, is what are the emerging technologies? What are you seeing out there in the field?

Shamala Sadananda: Life just changed for all of us on 22nd, November 2022. So everybody started talking about AI in much more mainstream than what it was years ago. I remember way back in 1996, I heard the headlines that says Deep Blue defeated Garry Kasparov in the World Chess Championship. And I was like, what sort of a name is this? And then we realized it was a AI chess program developed by IBM.

Kate Bevan: 1996. That's ages ago.

Shamala Sadananda: Yes. I was just checking. AI first came into being in 1956, and it took roughly about 40 years to develop the chess program. But since 1996 to 2023 now, it has undergone a sea of change. It has now evolved to something like a niche program to something where you're seeing mainstream adoption in industries. In our day-to-day lives, it's become omnipresent from smartphones to self-driving cars. Everything talks about AI. I know there's a model of a refrigerator, which actually has AI, and it can tell you that you're running out of milk so.

Kate Bevan: Do we really need an AI refrigerator though honestly?

Shamala Sadananda: That's where AI has gotten today. And coming to your question on trends, I see, of course, Generative AI, right? Everybody, that topic has stolen every headlines every day. But I wanted to highlight that Generative AI mostly, that the common belief and the terms and the news that you're looking around is where it is generating text and code. It is creating more of content. It's summarizing content. It's able to make a pro-grammer's life easier. You can change it from legacy to modernized code, but particularly what I'm seeing is multimodal AI models. Right. So what you're seeing is multiple types of data. It's not just creating new and realistic content for text and code, but images, videos, audios, music, 3D models. So all of those are particu-larly interesting because those were challenges that we couldn't have solved in the previous generations of AI, and that has been one of the top trends and top technologies that I'm seeing coming up in the industry today.

Kate Bevan: So I just want to pick one question before we move on because I hadn't realized AI had been around for such a long time. What were we doing with AI in the 50s and 60s?

Shamala Sadananda: I think it was used for computing and trying to make algorithms and mathematical models much, much bet-ter. Alan Turing trying to create the first computer in that context. So all of those things were trying to make it better. Most importantly, it was not as powerful and efficient, Kate. Right. It needed a lot of time to calculate two plus two is equal to four or break a code. So I think during that 40 period time, technology has also evolved to bring computers and memory and storage, all of that added to making AI possible. To-day, the most important and high memory component that we have is in our pockets. It has RAM, it has ter-abytes of storage. The other day I was looking at Costco and they have a storage, which is like a nano chip of three terabytes. I remember carrying hard disks to store information. So I think data has become easier, storage has become easier, cloud storage has become easier, and all of this has added to making AI possi-ble for common man like you and me, women.

Kate Bevan: Yeah. So it's been an iterative process, hasn't it? It's been a sort of building on the past things and all the things coming together. I'm old enough to remember watching the moon landings in 1969, and I remember my dad saying to me quite recently, he said, "The computers that we use to send men on the moon are much less powerful than the phone you carry in your pocket now." So I suppose AI is kind of the logical outcome of that. But what do you think was the thing that drove the leap into Generative AI? Because that's been a real explosion. I know it's been a long time coming, but what drove the leap?

Shamala Sadananda: I think it was all in that paper by Vaswani that came up and which helped build a more stronger approach to natural language. Creating those models and making it sound like you and me talking English became more important. But then these large language models started having more and more parameters. Being able to now take it to a multimodal AI, like I was saying, will solve real life problems. Right. Windows 12 will not have a start button. It will have a launch to copilot, which is basically something what Shamala and Kate want. Right. We start our day. I need somebody to tell me what I've missed during the last eight hours that I was sleeping. What was the top news? What do I need to do today? And also sometimes do it for me. I missed a meeting. Okay. Tell me what was the summary of the meeting?

Kate Bevan: Yes. That sort of brings us to what are we going to do with all this AI? What are the practical applications of it?

Shamala Sadananda: I am amazed to see how people are leveraging it. I mean, the news that I'm talking about that Generative AI is actually taking over is all about the applicability of these things. I was reading up something about dis-ease diagnosis. It is able to actually now read through multiple brain scans and areas of the brain. It can now diagnose what the problem is. So apparently there are tools which can help radiologists to interpret these brain scans and they can accurately identify that are brain hemorrhages. So I think people have start-ed taking AI to every walk of their lives. Nestle is trying to use AI to create virtual ads. So what you and I are going to see about the chocolate bar you're going to eat is a virtual ad that they created through AI. So I think it's trying to reduce time to market. It's trying to reduce effort that's taken to record these ads. It's al-so helping AI become more mainstream. I don't think Kate, AI will completely change what we do or com-pletely take over what we do.

Kate Bevan: What does this leave for the humans to do?

Shamala Sadananda: That's something we asked Microsoft in one of our conversations. Is it going to take away everything that we're doing? It's very scary, right? But if you look at it, it's not taking over. It's becoming your companion. It's becoming your co-pilot on everything you do. So a Shamala and a Kate now become more efficient and can do their things better. I think it's not going to completely change over and now doctors won't become robots. It's just going to make a doctor's job more efficient and more, I would say easier to use AI to save lives.

Kate Bevan: You see a lot of startups. What are the startups working on? What's the interesting stuff coming out of that ecosystem? Because we've talked a bit about what Microsoft is doing, and we know we can think about what Google and Meta are doing in that space, but what are the startups doing?

Shamala Sadananda: What is amazing Kate is there are roughly about last when I checked it was about 50 to 60,000 AI startups across the world. What these researchers are trying to do is use older models and create more powerful models or create new AI models and techniques to solve problems. Right. So for example, there is this startup as part of our Infosys innovation network called deepbrain.ai. They generate videos with AI. They do lip sync in minutes. There's another partner called Rephrase.ai that converts text to engaging videos in mat-ters of minutes. It's like you have existing people with their models available. I give them text and I click generate and it converts the text to the video and syncs up the lip to so perfectly. Right. It will feel like the person is actually speaking those words. There are startups like Digital Humans and Soul Machine, which I believe will be the future of customer experience. Right. They bring in animated digital tools, digital people.

So basically your existing conversational AI chatbots that we created over the last five, 10 years, now will have digital person delivering those words. Right. So to feel more personalized where I'm actually talking to a person with the lip sync and will speak the same chat components. Right. They have integration with Az-ure AI, CoreAI, all the GPT models. Interestingly, there's a startup called as Resemble AI, and they have a platform that creates custom AI voices. So I was thinking all those camcorders and video cassettes that I have and those memories that I have from my father, can I use those voice to generate his voice again, say-ing, hi Shamala, how are you doing? I would love that emotion, right? Amazing. I just feel if voice and cus-tom voices, custom people creation becomes like a core mainstream. Now we can create the voice and models of people who have deceased.

Kate Bevan: Digitally, it's a bit creepy though, because I've seen people talking about that and I've seen sort of some demonstrations of that kind of thing, and it's actually really unsettling. Should we be thinking about bringing back the dead? There's a whole sort of ethical thing there, isn't it? And also there's a sort of an ick factor about it as well.

Shamala Sadananda: You are right. So something like what Tom Hanks talked about recently, there was some insurance ad that used his voice and he had not authorized it. So I think you're right. The word ethical means a lot. We need to be responsible and we need to say that this is not any negative consequences of what we are trying to do, but somebody like my father or somebody who's close to me, I think it should be a decision taken by a trust or authority which, or gives the permission to leverage these voices. Right. So we were talking to a client who wants to bring back their founders pictures and videos, and he passed away in 1950s, but if he was somebody who could give a message, right, I think it would be impressive to me. That is a conversation I see on LinkedIn, I don't know if you've seen that Kate, is like Socrates and Bill Gates talking to each other.

Kate Bevan: See, I find that really unsettling and it feels like pushing the boundaries of what's acceptable to do with AI for me, because there are so many ethical considerations bound up around it. I mean, in your work, how are those conversations around ethics going? What are you seeing in startups and in the wider ecosystem?

Shamala Sadananda: And Infosys as well as all the startups have taken that up as it has to be a responsible by design approach, and we have to ensure where is the information coming from? We explain that. I don't think these decisions are being taken easily Kate. These decisions are being taken with the right regulators. The government is coming in now. I know EU is recently planning to actually by the end of the year, have certain laws. I heard about something where some executive orders are coming up in the next few weeks from White House as well. So I think regulators will come in. There will be the right responsible by design approach. What I am more interested in is to see the art of possible with AI, and that blows my mind. Bringing controls is defi-nitely one thing, and you don't want to do something spooky. But at the same time, I want to be sure that the technologies here and now and how we use this design and how we use this models is up to folks like you and me and where we want to use it.

There is this startup who can convert text to 3D models and Google AI had something like this with DreamFusion, which actually is able to say, put a kimono on a squirrel and it can actually create not just 2D images, but a 3D model. I had this question come from a manufacturing client 10 years ago, roughly. At that time, creating a 3D model or doing an image detection of a broken equipment or a component, doing the recognition was not possible because the component would have been broken. It would have a lot of dust on it, all of that. Right. But now if I can generate a 3D, I have solved a much closer problem to clients. So I really like how these startups are picking up their own niche technologies and changing the world in small bits and pieces.

Kate Bevan: That's a great way of pushing it. Do you find that the startups, because I think of startups as sort of move fast and break things. Are they very alert to the ethical problems and the secure by design and ethical by design approach?

Shamala Sadananda: Even if they're not, we should be. That's how I see it. People like enforcers who are building solutions and platforms are taking an AI first approach to things where we are taking those startups to solve bigger prob-lems. I feel we are the ones who should bring in those ethical models or ethical needs to these problems, even if they don't do it. That's our approach, the way we have taken forward to some of these solutions.

Kate Bevan: So you'd say it's sort of down to the leaders in this space to make sure everybody follows along, and partic-ularly I suppose, to the vendors and the services providers?

Shamala Sadananda: Rightly. But Kate, you said it. I am sure there will be regulators that will do this better, but until then, it's up to you and me.

Kate Bevan: That's both a good thing and a worrying thing. Oh, I think this brings me to my next question. It's a way of sort of framing what other things you're worried about, but do you think the AI is going to kill us all?

Shamala Sadananda: No, I don't think so Kate. I strongly believe it's going to save lives. It's not killing us. It's not killing mankind. I worry that if we hand over every bit of it to AI and it starts learning by itself, I think that's the worry every-body has. That's the fear. When it starts making decisions for us, even without me telling it to do it at that point, it will start worrying everyone. But I think today, if we control it, we will still be able to use it for the best of our needs. It can create medicines. It can find diseases. It can do so much. I know that it is contro-versial and there's a lot of talk about ethical and aspects to things, but if you use it the right way, I think AI will save lives.

Kate Bevan: That's great. That's really optimistic. Thank you, Shamala. That's fantastic. Thank you for joining us.

Shamala Sadananda: Thank you for having me. It is always a pleasure.

Kate Bevan: That was part of the Infosys Knowledge Institute's AI podcast series, the AI Interrogator. Be sure to follow us wherever you get your podcasts and visit us on infosys.com/iki. The podcast was produced by Yulia De Bari, Catherine Burdette and Christine Calhoun. Dode Bigley is our audio engineer. I'm Kate Bevan of the Infosys Knowledge Institute. Keep learning, keep sharing.