Subscribe
Share
Share
Embed
upGrad Enterprise aims to build the world’s largest GenAI learning initiative to enable high-growth companies to embrace technology’s transformative business impact. Hosted by Srikanth Iyengar, CEO, upGrad Enterprise, the GenAIrous Podcast, will curate an exciting roster of global experts and guests, who are at the cutting-edge of Generative AI, and its varied applications in the world of business.
Welcome to the GenAIrous Podcast where we unravel the fascinating world of generative AI and its transformative impact on business globally. I'm your host, Shrikant Iyengar, CEO of upGrad Enterprise. At upGrad Enterprise, we're building the world's largest Gen AI learning initiative, empowering high growth companies to leverage cutting edge technology. Each week, join me and the roster of global experts as we explore innovations shaping the world of work as we know it. Let's get GenAIrous.
Srikanth Iyengar, CEO upGrad Enterprise:Hi there, and welcome to another episode of the GenAIrous Podcast. I'm delighted today to have with me Rahul Singhal, Chief Product and Marketing Officer of Innodata. Rahul joins us from Short Hills, New Jersey. Rahul, welcome to the show, and thank you for sparing the time.
Rahul Singhal, CPO - Innodata:Thanks, Srikanth, for having me.
Srikanth Iyengar, CEO upGrad Enterprise:Wonderful. So, look, I'm just gonna get straight into it. InnoData is an exciting company. You're publicly listed. You're at the heart of the Gen AI transformation. So tell me, why is Gen AI a transformator?
Rahul Singhal, CPO - Innodata:Really interesting, question. Right? So, think about the journey of, technology. Right? So you had the the PC era, which was transformative. Then you started with the Internet era. Right? Then you had the mobile era. And then you had the AI era. Right?
Rahul Singhal, CPO - Innodata:AI was transformative. We started to do tasks which were which the patent recognition and machines could solve the problem. Gen AI is a different piece. Gen AI is today's technology is memorizing everything that humanity has ever written and is being able to predict and solve problems which humans could have never solved. And, if you look at companies like Google and Microsoft and Amazon and OpenAI and others, pretty much every CEO has said this is the birth new birth of the Internet, right, in which we completely subscribe to.
Rahul Singhal, CPO - Innodata:So, yeah, I think gen Gen AI is here to transform humanity as as a whole.
Srikanth Iyengar, CEO upGrad Enterprise:Completely completely agree, Rahul. We are seeing that too. I think, we're just scratching the surface, and we'll get more into it as we talk. But, let's get a little more into the the the technical aspects of it. Right?
Srikanth Iyengar, CEO upGrad Enterprise:Clearly, Gen AI impacts in multiple levels of the stack, whether it be the computing power or, you know, the algorithms that run on it. But, of course, you play in the data layer. So why don't you give me a sense of what it means in terms of data for enterprises? Because you deal with Fortune 500 companies every day. Yep.
Rahul Singhal, CPO - Innodata:Yeah. So if you think about, training or how these models get created, there are 3 layers to it. Simply put, there is there is a data layer to it, which is, machines are dumb, and dumb machines need data for them to understand and reason. Right? And that's that that's the data layer.
Rahul Singhal, CPO - Innodata:The second layer is compute, so you need large amounts of computing part that could be used to train these models. And then the third is the algorithm. Right? The algorithm that that are used to train the models, which is mathematics. Now if you think about of those 3 layers, algorithms are pretty much open sourced. Like, most most, companies are using open source models. And then fine tuning the word of fine tuning, which is adding their proprietary data to train those models. And computers, computer's a commodity. Right? So you can buy compute. You can buy it from NVIDIA. You can buy it from Korviv. You can buy it from Amazon. You can buy it from Google. Right? So at the core, it is the data that is critical to the success of any of these, generated AI models.
Rahul Singhal, CPO - Innodata:And we've been, Innodata as a company, just to give a, give a 30 second snippet on who Innodata is. So our, we've been around for 35 years, publicly traded. We are leaders in data engineering. And for the longest time, our company was, supporting information publishers and producing high quality data from unstructured data to structured data.
Rahul Singhal, CPO - Innodata:So we got to, we got to hone in our skills of how do you create data 99.995% accuracy. Right? So think about, think about a customer where you, where you you're looking to, you're looking to tag a data, which could be very nuanced. Right? Requires subject matter experts. Our company was doing; you're allowed 1 in 10000 mistakes. Right? So think about medical data extraction from medical records. You have we have doctors who are on our staff who are looking to extract data for underwriting purposes. The world of Gen AI now requires that expertise where you require people who can train that high quality data that is then fed to the machine.
Srikanth Iyengar, CEO upGrad Enterprise:No. Completely makes a lot of sense. And I think the, you know, golden datasets like we were talking earlier, is something that's critical. But let's talk about the skills. You know, you talked about medical data where you use doctors. Obviously, you're using domain specific people. But given what's happened in the Gen AI space in the last 2 years, has it changed the kind of skills that you recruit for?
Rahul Singhal, CPO - Innodata:It's a great question. Yeah. So, the what I like to tell our people is, the new, coding language is English. Our languages in general is the, is how code is written. Right?
Rahul Singhal, CPO - Innodata:So if you think about these large language models and there's data is out there, companies like OpenAI and, Microsoft and others, 40% of code is now written in English language. We are seeing some very interesting statistics. For example, our company just in last year have hired 1500 plus associates who are humanities major, right, across the globe. Now think about that as a...
Srikanth Iyengar, CEO upGrad Enterprise:You said 1500 just for our listeners. That's not 15. That's not 150. That's 1,500. Wow. Okay. Yeah.
Rahul Singhal, CPO - Innodata:Yeah. These are humanities majors across, like, people who have bachelor's in journalism, bachelor's in creative writing, bachelor's or master's in, master's in filmmaking. Right? We are hiring people who are, who are humanities majors, and we're also hiring people who are domain specialists. So we are hiring people who are, physics majors or who are biologists and geologists and agri techs.
Rahul Singhal, CPO - Innodata:So domain specialization and those kinds of people are being used because these models require to understand the language of that domain, and you need to have very pristine data that needs to be created. As jobs are getting transformed, right, there is going to be and there is we are seeing there is a significant need for upskilling of talent around the world in these new areas, which is going to going to pick up. Right? So because, you could be a bachelor's in English, or you could be a bachelor's in marketing. Right?
Rahul Singhal, CPO - Innodata:And you don't have a job. But there are jobs being created where, for example, for Gen AI. Just take Gen AI training, data market, right, which is expected to be a $7,000,000,000 market in 3 years. Right? Wow.
Rahul Singhal, CPO - Innodata:The amount of data needs the and this is the Bloomberg statistics, being published where, companies are gonna spend over $7,000,000,000 in 3 years just for training data, across different modalities. Right? You have image and video and text and, audio data, or multimodal data. You need you need training for associates who can be up and ready and be deployed on projects in a week's time. Wow.
Rahul Singhal, CPO - Innodata:And that is where the opportunity of upskilling and rescaling of talent is required, where companies are looking to not train, but but to be able to, use trained talent and deploy them on projects.
Srikanth Iyengar, CEO upGrad Enterprise:No. That makes a lot of sense, and thank you for the, you know, emphasizing the importance of upskilling. But coming back, you know, to that coding versus English or, let's say, humanities or, let's say, core domain skills, you know, you you you've done a start up yourself in the past. You're close to, you know, start ups in the Bay Area or elsewhere across the globe. And a typical, you know, thought for a start up would be, let's get a bunch of people who know code together and let's crunch code. How will start ups in the future evolve, do you think? Do you think will it be different skill sets? Yeah.
Rahul Singhal, CPO - Innodata:Yeah. I think it's a great great question for let's just think about just the future. Right? So I think we are probably 3 to 5 years away from a unicorn startup being a 1 person company or a 2 person company or a 3 person company. Right? Because these language models and now the the world is talking about agentic models. Right? Which is, which are small pieces of code that can do a specific task and take an action around it. So for example, let's take a agent that is being created for scheduling purposes. Right?
Rahul Singhal, CPO - Innodata:You have we we have assistants who help us and coordinate our busy schedules. I could have an agent, which is sitting on my laptop, and these are the kinds of agents that are not doing the work, which understand they understand my calendar. They understand, the email that is being sent, and they're they're communicating as as humans and saying, hey. Raul is available in, eastern time zone between 2 to 2 25. Would you have availability?
Rahul Singhal, CPO - Innodata:And they take action and create that meeting for me. Right? Lot of that work is now these agent workflows are going to append the work. So think about the start up fold. When coding you don't need coders and you can basically take you can create use language to create applications.
Rahul Singhal, CPO - Innodata:That's the world we're gonna be in. And then you can have agents who are doing SEO for you, agents who are doing HR, finance for you.
Srikanth Iyengar, CEO upGrad Enterprise:But let's talk about a few domains. We you know, you touched on the fact that different domains could get disrupted and maybe different functions, you know, whether it be finance. You talked about invoice processing, but even something like a marketing. Maybe take an example of a domain and give us a sense of what changes on the ground.
Rahul Singhal, CPO - Innodata:Yeah. So, let's take an example, within our company. So we have, we have a software for PR professionals. It's called Agility PR, and the software does three things. It does monitoring.
Rahul Singhal, CPO - Innodata:It does media monitoring. So we ingest, like, billions of news articles and we monitor for, news, and topics that are interesting for a PR professionals. We have a journalistic database where we have, we we know a journalist what what they write about, and have data on how they write about. And then we have tooling that that's available to a PR professional that can, that can write within our tooling and then send it to those journalists so that they can pick up a story and then, you can, you know, do PR, PR around it. We, we created our we fine tuned, a large language model using, the data that we have around how an a PR professional writes.
Rahul Singhal, CPO - Innodata:And today, a story that used to be written by a PR professional would take them 3, 4 hours. They can now write a story in a minute. Literally a minute. You basically tell them, I want to write a story about generative AI, and then my new, application is about, how I can, how I can do better SEO. Right?
Rahul Singhal, CPO - Innodata:And I wanna pitch it to these 10 journalists, then this is my value proposition and this is an area where you can pick up my factoids. Right? You basically just give it those inputs. It will create 10 different versions of, of a story that can then you can edit to pitch to different PR professionals. Imagine this, what it does. It has created, like, almost saved 90% of time for a PR professional today, and they can do a lot more and reach out a lot more and then hopefully drive lot more practice. So that's the transformation that is happening.
Srikanth Iyengar, CEO upGrad Enterprise:And yes. And that will happen, you know, domain by domain, industry by industry, completely. So now tell me about your clients. I mean, like, we, you know, talked about before, you work with some of the biggest companies in the world. And, you know, nowadays, you know, you you know that most companies can't even speak for 5 sentences on an earnings call without mentioning generative AI. It's clearly top of mind for CEOs, for boards, for divisional executives, for leadership teams. But what are you hearing or seeing clients do? What does this really mean for them? Is it something they just have to do? Is it something that they want to do? What's the mindset?
Rahul Singhal, CPO - Innodata:I would I would break the clients into 3 segments. Right? So I would break them into builders. So they're companies which are tech companies, which are building foundation models. Builder would be a company like an OpenAI or Meta or, Amazon or host of startups, which are saying, hey. I wanna own I wanna create a model that is then will be used by enterprises, from an implementation perspective. Right? So there are found builders and there are adapters. So I think what we are seeing today is technology companies, I would say fortune 200. Right?
Rahul Singhal, CPO - Innodata:Mid, which will include pretty much every, every large tech company are building their own foundation model or a large language model. Right? So and it's an arms race. It's an arms race. Everyone is worried to be left behind.
Rahul Singhal, CPO - Innodata:So we see enterprises doing lots of POCs and, Srikanth, as you rightly said, it's a C level, board level, conversation saying, you talk to every CIO and I talk to CIOs all the time. Like, their leadership is saying, what are you what are you doing with Jenny I? Can you tell me how you're implementing it? Because investors are ask asking for it. We are seeing marketing use cases, the things like content marketing, the PR use cases, though those are going in production.
Rahul Singhal, CPO - Innodata:We are starting to see slivers of, HR use cases like onboarding. The things which are, which are easy enough getting onboarded and internal q and a use cases. Right? So knowledge management use cases where, you can create a q and a chatbot for for internal purposes to drive, better, better results of productivity benefit. Right?
Rahul Singhal, CPO - Innodata:So efficiency use cases are what we are seeing. So I think that world is going to change over the next year, 2 years as the technology is mature, the cost of compute and cost of models comes down. We will start to see digital transformation at an enterprise level happening where they will start taking large scale prod, product workflows and transforming it or infusing the LLMs into there, which is the large opportunity which we, as a company, see is probably going to be a $1,000,000,000,000 opportunity.
Srikanth Iyengar, CEO upGrad Enterprise:I completely agree. And in fact, all of the processes or the functions you talked about, whether it be call center or customer service, marketing, HR, internal productivity. We see all of these use cases. In fact, I'd probably even add enterprise sales because, you know, sales has traditionally been an art, but now tooling that so that people can prospect better, close better, understand the customer better is another area that's getting disrupted. So clearly, agree with all of that.
Srikanth Iyengar, CEO upGrad Enterprise:But yeah. You know, like we touched on earlier in the conversation, we're just scratching the surface. We're probably not even 2 years into it. November 24 will be 2 years since GPT was released. So really, it's very early stages. So in your sense, what is the 5 year view? What's how is this space gonna evolve? What's gonna happen in your view?
Rahul Singhal, CPO - Innodata:I think that we are here for we are very, very early in this journey. Right? So I'll talk about why, what's going in our view, what how we are betting and transforming our business, for the next 10, 15 years. So if you think about these model, there are, 3 components to a model. Our first component tends to be language.
Rahul Singhal, CPO - Innodata:Right? So today, if you look at chat GPT, it works in all these languages, but predominantly it's been trained on English data. Right? Now you're going to be if you think about the world to date, it has 226 languages plus thousands of dialects. You need the models to be trained in all of these languages at scale, which will require a globalization, globalization of, of training data.
Rahul Singhal, CPO - Innodata:Right? Which will which can then be used to train the model. So there is a language component and globalization component that has to happen. Then there is task. Right? So if you think about what the aspirations of these models is, these models are being trained on tasks like a q and a. I can do question answer. I can do summarization. I can do video editing. I could do video captioning.
Rahul Singhal, CPO - Innodata:So you need training across task and imagine the task that we as humans do. There is innumerable number of tasks that that we as humans do, so they have to be training that will happen across that. And then there's the domain. So you have to train it on a physics and chemistry, and it's not about legal. It's about contract laws, very different than privacy laws, then very different than real estate law.
Rahul Singhal, CPO - Innodata:Right? So you will need training of these models at the subdomain level. That's the and then don't forget these models today are memorization models at best. They we are just starting to see reasoning models, which is the OpenAI Strawberry, 4.0 model that got released, month ago, which is very early. It's first very, very early stages in how a model can do deduction and reasoning.
Rahul Singhal, CPO - Innodata:Right? And I think that will 5 to 10 years later, we will start to see these models be able to reason, but there it's it's ways from there. So I think there is so much transformation and so much, opportunity in this space, that we as a company are very gung ho about it.
Srikanth Iyengar, CEO upGrad Enterprise:No. Absolutely. And look, the way you, you know, sort of portray that picture, 3 vectors or 3 components, language, task, and domain, You know, it almost reminds me of the start of the Internet, like you said. Right? When the Internet age came, people are trying all kinds of use cases, all kinds of ways. And, yes, there were mishaps, there were failures, but the world changed irrevocably for the better. One would like to think largely for the better. And I think that's what's gonna happen here as well. You're gonna have multiple ways in which models evolve, which lead to value creation. Many will work, some will not. People will experiment. But but, you know, clearly, in the longer term, we will progress. So completely could couldn't agree more.
Srikanth Iyengar, CEO upGrad Enterprise:Raul, this has been a fascinating chat. Sounds like, you know, you're at the, you know, at the cusp of what's a very exciting journey, both in terms of the trends happening in the world, but also, you know, data is in a very exciting space.
Srikanth Iyengar, CEO upGrad Enterprise:Like, we were, you know, talking about earlier, you guys have been in the business for 35 years. You personally been in the AI space for probably 11, 12 years now. So clearly, you saw something the rest of us didn't. You knew Gen AI was coming?
Rahul Singhal, CPO - Innodata:No. I wish I did. But, hey, I'll take, I'll take luck over, over anything else, over intelligence any day.
Srikanth Iyengar, CEO upGrad Enterprise:Yeah. I think it was Napoleon who said, "I would love Smart Generals, but if I had a choice, I'd pick Lucky Generals over Smart Generals any day." So agree. But clearly, you're in the right space. Listen, all the very best to you guys, and thank you for joining us today.
Rahul Singhal, CPO - Innodata:Thank you, Srikanth, for having me.
Srikanth Iyengar, CEO upGrad Enterprise:And that concludes another episode of the GenAIrous Podcast. We are very grateful to our guests for their time and expertise. A big thank you to our producer, Shantha Shankar in Delhi, and our audio engineer, Nitin Shams in Berlin, for making magic happen behind the scenes. Join us next time, and don't forget to subscribe to GenAIrous wherever you listen to your podcasts.