AI Interrogator: Generative AI Adoption Across Industries with Mona Dash

Kate: Hello, and welcome to this episode of the Infosys Knowledge Institute's podcast on all things AI, the AI Interrogator. I'm Kate Bevan of the Infosys Knowledge Institute, and my guest today is Mona Dash, who manages AI and automa-tion practice sales for Infosys in Europe. Mona, thank you very much for joining us.

Mona: Thank you for inviting me.

Kate: Well, it's great to have you. Since you see a lot of what's going on in companies, can you just give us a sense of what you're seeing in companies in how they're deploying AI? What AI technologies are they looking at? What are they using? And also, how are they using it?

Mona: The first thing of course is we are literally seeing a lot, and I mean really a lot of interest from all our customers, irre-spective of vertical or respective of where in the globe they are.

So yeah, I mean, just as a background as well, if you think of it, AI as a term, it was actually coined in sort of 1956 in a conference by John McCarthy who's considered the father of AI.

Kate: That's even older than I am.

Mona: Yeah. Well yeah, well, I mean, it's long back, as if 1956. And it's a conference in Dartmouth in the US. And between then to now, into 2023, AI has been sort of there in our lives, but it's only been the last six months that things have been happening at very fast scale.

So in terms of AI technologies that our customers are asking for, I would say it's all a lot around generative AI. So the traditional AI, I mean ML modeling, cognitive automation, intelligent automation, we were always used to that, but the last few months have been literally about generative AI.

Kate: So are you seeing it so that they are playing with consumer versions of generative AI like ChatGPT or are they doing something else?

Mona: ChatGPT of course has a very consumer-related use cases and they are very conscious that it's not really enterprise-ready because ChatGPT didn't exactly come with APIs initially as well, so they are experimenting more with the en-terprise versions which are more of the other GPT models.

We like to say that we have kind of seen some broad adoption patterns, and we actually see that across almost eight areas. But if I have to mention the top three, I think it's a lot within the software engineering space where they're looking at use cases around helping developers basically doing coding-related tasks or test-related work. So whether it's code generation or test case generation, so things like that.

And the secondary of course is the whole space of customer experience and customer service, because as you know, ChatGPT can answer questions in a way more effective way than something like a traditional chatbot did.

So it is also augmenting their existing conversational AI platforms with the generative AI capabilities. So that's some-thing which is a big space around the whole customer service area. And doing things like semantic search as well, which can help with questions being asked and basically giving more contextual answers.

And I think the third area as well is around contact centers, help desk, or even sort of things around business opera-tions and IT operations.

Kate: So it's almost like the user experience is both external and internal?

Mona: Absolutely, yes. Actually right, because employee experience is also something which can be helped, and we are doing this within Infosys, we have a lot of drive about being an AI-first organization. So within Infosys itself we have the whole vision of having an AI buddy for every employee. But currently we have a lot of work going on in the sort of code assistance or proposal assistance or things like that, yeah.

Kate: What sectors are really going all in on it? Is it creative industries? Is it healthcare? Is it life sciences? Where are they going all in?

Mona: I don't think there's any particular vertical which seems to be leading more than other, right? So it's across all verti-cals. But probably the highest, I think, activity might be more within FS and manufacturing and CMT like telcos.

Kate: That's interesting. We just tend to think of financial services as being slightly ahead of the curve when it comes to adopting new technologies because every tiny bit of advantage is really important there.

Mona: Absolutely. Yeah.

Kate: What are they getting right when they're doing this? And also, what are they getting wrong, do you think, when they're doing this?

Mona: Most enterprises I think are being quite cautious, which I would say is probably the right thing to be. What I mean is generative AI obviously.

So they are being cautious in the terms of what is the sort of security they need to be aware about, data privacy, what's exactly happening with the data? They're also conscious about how they're rolling it out in the organization. And they want to get it right, they want to get the right framework in place before actually doing much more or be-fore kind of experimenting too much.

And it's almost like a flashback to the time RPA came into our world and which was sort of early, I think between 2003, 2005 or so, a lot of RPA product companies were set up. And in the last decade especially we saw there was huge excitement around automation and how automation would take over our jobs and whether robots are taking over the world. And in all of that, RPA really manifested itself very quickly and there were companies which were adopting automating at a very fast scale, but a lot of governance and COE frameworks actually got set up later.

Kate: So this time they're out in front of it, are they?

Mona: Yeah, I think it seems to be that this time they're trying to understand things almost bottom up, get a better visibility of the business case. They're really trying to understand that, "What exactly is this technology? How do I put it out across the enterprise?"

And maybe because generative AI came in this very democratized form because it just is so accessible to everyone. So I think because of that, they're very conscious that, "How do we actually roll it out in the enterprise? How do we just make sure that we're getting the right benefits about it?"

Kate: One of the things I've been looking at is the use of synthetic data to kind of pad out training data. Is that something you're seeing? Are you comfortable with the data I think is what I'm asking here?

Mona: All AI M&A is always dependent on the data, of course. So whatever the data's fed in is what the outputters get. But I mean, I am not obviously a data expert at all, but there is definitely a huge correlation there.

But the point about this whole large language model, since they have been trained already on such vast volumes of data so the output can be quite good. I mean, obviously we have to still do prompt engineering and fine-tuning of the output, but the large language models itself, they have been trained literally about on everything in the world. So yeah, so I think that might take away some of the problem I think which was there in traditional situations.

Kate: Coming back to the flip side of that question, what are they getting wrong?

Mona: I think, I mean, one of the things we like to say is that if you have a hammer, then every problem looks like a nail. So we do sometimes have asks and requests that let us solve this problem because we want to use generative AI.

And often the problem may not be solved best by generative AI, because the point is generative AI is huge possibili-ties, but it's also not a magic wand. I mean, there is obviously infrastructure costs, there is sort of support costs and all of that which comes in, so maybe some of the problems can be solved by a more simpler or a different model.

So we do like to advise and consult from a sort of holistic point of view. So we do want to then discuss the possibili-ties which are available and also discuss the art of the possible.

Kate: I suppose then that you are kind of helping people not be overwhelmed by the possibilities of it?

Mona: Yeah, I think as in our role in Infosys, I think we are very conscious that if we don't want to overwhelm, that the pos-sibilities are huge and you can literally reimagine everything. But it's also that how do we sort of move forward, let's find the right technology and to solve the problem, and then maybe even look at redefining the problem statement, because generative AI could actually probably bring in a huge amount of re-imagining.

Kate: So this is actually a real watershed moment for businesses right across all the sectors, isn't it?

Mona: Absolutely. Absolutely. I think it's not something which has been seen before, because if you look at curves of other different technologies and how much time each took to actually kind of grow on the trough of disillusionment, as I'd said, generative AI I think, it's kind of one of the fastest growing ever. And I don't think anybody actually saw this coming. Literally from November 2022 I think is literally when the massive boom is supposed to have started off from with ChatGPT.

Kate: Mona, one of the things we talked about when we discussed this earlier was actually that you are in your own time really creative. And I'm just wondering, as a creative, what learnings do you take from your work with AI into your creative space, your personal creative space? And also interestingly, what do you bring back from that into work?

Mona: So the first time I think my worlds are sort of intersecting, and there's a lot of things about how generative AI is actu-ally bringing both the left and right brains together, which has not often been seen because traditionally I think eve-ryone's roles have been very different, the creatives and the technologists.

I mean, in terms of how I normally manage this, I basically write in my personal time. I bring the creativity in the work in the sense that how you approach a problem, right? You kind of bring in some more of maybe emotional empathy or how do you approach the problem in a more creative way rather than a structured data way? And I think this is how generative AI is actually helping us as well.

So that is one thing, but I have to also mention this, that there is use of generative AI in the creative space. And I think that's something which I feel that has to be more looked at or more explored because there's almost a situa-tion that the creative process itself, and I'm talking about musicians or authors or artists and painters, because obvi-ously generative AI can generate any kind of content.

So I just feel that this is something which needs to be looked at because in fact, there's been a thread going on where some 10,000 or more authors from The Authors Guild have written an open letter to OpenAI and AWS and Microsoft and everybody to compensate writers for the copyright they're using, because literally everybody's works have been put into training the AI models. So you can find Dickens, Shakespeare, everybody. That I think is a thing which needs to be really looked at.

Kate: And finally, I've asked this of everybody, and it's quite a provocative question, but I think it's a good way of getting to some of the issues, do you think the AI is going to kill us all?

Mona: That's a good question. So, and I think as human beings, there's been so much of science fiction and movies and books across the time when everybody imagines that the bots have taken over and the humankind is all lost.

So the point is, all those times, even with RPA, like I said, that had been concerned that bots will take over, but bots always, they did things which they were told to do, and it's for the first time that we actually have something which is generating its own content. So I just think that this is incredibly powerful, more than what we are probably used to.

So I know, and I don't know if you've heard about this, but there was an research which OpenAI did when they actu-ally launched GPT-4 or they were testing GPT-4. Apparently GPT-4 was smart enough to solve the CAPTCHA prob-lem, because what it did, it actually tried to kind of get a human to take it through a CAPTCHA. And when it was asked that, "Are you a robot?" It said that, "I'm visually impaired." So it sort of seemed to make up a lie in that sense.

So there are these kinds of things trickling in, because is there a way that they're actually going to be generating orig-inal thought? And I think that's where the sort of risk is.

So there are all these sort of doomsday, of course, ideas that five ways generative AI can kill us all. And equally, there are five ways that generative AI can save the world, and I think we sort of focus on the best things that tech-nology can really do, and there's huge benefits which it can bring. So we just need to have the right guardrails, the right intention, I would say, and the right ethics to go about using AI.

Kate: I think that's great. And thank you so much, Mona, for your time and your thoughtful answers. Thank you for being on the podcast.

Mona: Thank you for inviting me.

Kate: That was part of the Infosys Knowledge Institute's AI podcast series, the AI Interrogator. Be sure to follow us wher-ever you get your podcasts and visit us on infosys.com/iki.

The podcast was produced by Yulia De Bari, Catherine Burdette, and Christine Calhoun, with Dode Bigley as our au-dio engineer. I'm Kate Bevan of the Infosys Knowledge Institute. Keep learning, keep sharing.