Dr. Vishal Sikka founded Vianai Systems Inc in 2019 with a seed investment of over $50 million and an advisory council of some biggest names in enterprise technology and academia. Prior to Vianai, Sikka was CEO and Managing Director of Infosys and CTO and executive board member at SAP. In addition to serving as founder and CEO of Vianai, Sikka sits on the boards of the BMW Group, Oracle, GSK and on the advisory council of the Stanford Institute for Human-Centered AI. In June 2021, Vianai closed $140 million in Series-B funding with SoftBank Vision Fund 2, who joined existing individual investors. He spoke to Fortune India about AI, his plans and how India can be relevant in the field.
Edited excerpts:-
Q. You have a Ph.D. in AI, which many may not know, and are now betting on its potential dominance for the future?
A. When I was looking at the next 20 years, it was clear we are entering a different kind of phase in the technology industry. No IT department can come close to competing with the big clouds. I felt the transformative potential that AI has, is basically as big as back when computing started in business in the 1970s and 80s. So, I went back to the drawing board to see what the important AI problems were and that challenged us. That guided us into what our first products were going to be and we started building those products over the course of those two years. Also, we went to raise our second round of $200 million, bringing in SoftBank, in addition to our existing investors.
Q: How close to human intelligence is AI?
A: If I step back for a second, there is a lot of hype around AI but one thing that is very clear is that what the world today needs is human-centred AI. The idea that AI can replace human judgment, or that automates humans out of business activities completely, is wrong. Basically, AI is very good at amplifying human judgment, in such a massive way that we never imagined.
But many elements that we take for granted in our day‑to‑day work, like without human judgment, common sense, the ability to make sense of things, and understand things done with exception and wrong. We are simply nowhere close to doing the same. I am on the board of BMW and we keep debating these autonomous cars, we already have L2 and L3 abilities of autonomy. Within this decade, we will also have L4 autonomy, which involves, autonomous driving, but in a very controlled kind of an environment, with the closed-loop system but L5 autonomy, which is basically wherever you walk into the car and you go to sleep. The car that will take you wherever you want to go will not happen even 50 years from now.
Q: So where does data analytics stop and where does AI take over. Are they one and the same. Or is there a blurring line between the two?
A: There are several ways to look at it. The reality is enterprises AI today is basically extensions of the data world. If you take analytics beyond, it is an assessment of what happened, and what took place, you have data and you run questions on it, and you get an answer to these questions. If you take that a step forward and start to look into the future, what is likely to happen, then you get into things like where you try to observe patterns, learn those patterns into functions, which is basically what machine learning is and then you predict what is going to happen. Generally statistical machine learning, or even neural networks and so on, are the mechanisms for us to learn from data and craft functions out of data that can help us answer much more complex questions. In that sense, you deviate from analytics and go beyond, into creating, and learning from data.
One of the pioneers of our field, called this ‘curve fitting’, where most machine learning today is curve fitting and if you take a whole bunch of data about something like, what is fraud in a bank, or should you approve this transaction or not, or is this machine going to fail or not? So you look at data and then you fit that to a function that predicts for you that under these circumstances, this is a fraud, and this is not fraud, or this machine is going to fail or not and so on. And generally, these techniques are no longer explainable techniques.
Today, generally the technique is to throw a whole bunch of data at a machine learning system and it builds on the data, like when a car recognises obstacles on the way and so on, again, it is basically exercising deep neural networks that have been trained on what is an obstacle and what is not. But there is another very important sense in which AI goes beyond data and now I'm getting a little into philosophical terrain. It is a sense of embodiment. So, if you look at one of the criticisms of what is called a large language, models like Burt and recently, Google launched PaLM and so on. These are really massive neural networks that are comprehensively large. PaLM has 540 billion parameters and those things, perform with deadly accuracy if you're going to ask questions about anything and sentence completion, entity, extraction, or summarisation. But ask a new non‑trivial question, and then you realise these things are still very far away from human comprehension.
So the embodiment problem is when we learn a language, when the child learns language, the child is an actor in an environment. We have senses, we interact with the environment we give and we take, and that act gives us the ability to learn and respond. We don't do this in the systems today. So one essential characteristic of AI is that there is an embodiment.
You might have a simple application, it could be a facial recognition application on your phone, but there is an embodiment that interacts with the outside environment.
And so in that sense, those applications that apply AI towards where there is an interaction involved with the outside environment, are far away from the world of data. So you can think about it from that point of view, but the simple answer to your question is that, by and large, most AI in enterprise are all data-oriented but will change over time.
Q: If we take that thought a little further, AI is already transforming many areas of commerce, even service areas (Law firms etc., ), but can that also change the fields like music, art, and literature?
A: I serve on the advisory board of Stanford's AI centre and we have some professors who look into how domains like art or dance and medicine get impacted by AI. The reality is that they are quite far away from human capability. I mean we have sense amplifying systems that amplify our senses’ perception. So, earlier we talked about data, and the reality is the data has footprints of some activity.
There was a person who was looking at a website, trying to make a purchase, and then decided to not make it and went away somewhere. Why did this happen? Who knows why this happened? All you have in data is the footprint of this activity of the person.
So the purpose of building a model to capture something is to try to understand the reality behind the appearance. I give this example the first time that they discovered the binary star (two stars spinning around each other), the second star was not visible, but they could tell that there was another star there based on the fluctuations, the orbit of the star.
And because you understand the laws of physics you can say, oh, wow, there is that shift happening, which must mean there is a star we can't see. So trying to infer the reality behind the appearance… none of these systems can do that right now. We are still very far.
Q: There is a slowdown of sorts expected in technology in the near future. Is that a concern?
A: No, not at all. It is the best time to build something innovative because you don't need innovation when times are good. It’s needed especially when times are bad. I saw that happen in the aftermath of the dot-com crash, in the aftermath of the Lehman brothers and the 2008 financial crisis. It always happens. Did a great set of companies emerge out of this? Yes, and in the future, I hope that mine will be one of these, and help transform the world for the better.
I think the potential of AI to help improve our operations, help improve customer engagement, to help come up with new products that are better, is really an enormous one. I think India still has enormous potential in front of it and every time I speak to the Prime Minister (Narendra Modi) I emphasise there is such potential in creating a digital future for a billion people who are AI literate, who are computing literate, to build innovative products for the outside world. You know, Reliance Jio for example, is starting to go international and make offerings outside India… It is one thing to provide people but quite another to be building products for the world. The sky's the limit when it comes to India.
Q. Tell us about the products you have been building?
A. We just launched our first application yesterday. So this is an area of customer engagement, which is one of the most important areas for our business, and the pandemic accelerated customer engagement in a really dramatic way. Every business more or less became a digital business, or basically, overnight, even the ones that were sort of on their journey, but not quite they literally had to embrace it. So what we found was that, with certain advances in AI that we built over the last three years, we can really revolutionise the customer experience and engagement. We have a team in Israel, led by Jack Klein, which has built this product, which we now call revenue science, it applies causality causal inference. So this causality AI is one of these new frontiers in AI.
Especially, over the last few years, there has been a lot of work in causal inference, separating correlations from causation, what causes things to happen? It is now possible with these techniques to look at data, run large amounts of machine learning, experiments on the data and identify causal relationships.
And these can be applied to improving customer engagement in a very purposeful and dramatic way. We had great success with some of the largest companies in the world applying this. While building this product, we needed certain capabilities from the point of view of data. So, we acquired a small company that gives us those capabilities. It's a company called Deal Tale, and we launched this product for our 40 to 45 customers. Generally, these are technology companies, about half of them are unicorns and there are some big companies as well, Net App for example.
Q: The world economy is in a state of flux. So is it a challenge in building an AI firm. Now that you’ve raised a couple hundred million, what will you deploy the capital toward?
A: I have learned that crises are very good times to build a company. It clarifies things, makes things more real and it is a good time to be an entrepreneur. I was keen to raise money because you kind of saw this coming. There is a buffet ratio, which is the ratio of the market capitalisation, to the GDP. And we were already at this ratio being at a historic high and you could see that what is happening now was inevitable. I'm not concerned about it, but yeah, there will be a lot of clarifying that will happen in the market in the next two, or three quarters. So we are fortunately in a good position. We are applying capital primarily to the resource, R&D and finding talent in our case, we also have to build our own. As an enterprise company, we have some of the largest companies in the world as our customers. So we have to build a sales force, and that is, not a cheap thing to do and so on.
We just crossed a hundred people, and are a company of 105 employees, today. About 25% to 30% are women and roughly the same number of Ph.D.s. So one of the things that we have done is we have gone with the strategy of hiring people wherever they are. Because of the scarcity of talent and because of the need to hire the right kind of talent. We have operations and about 35 people in Israel, we have a few people in Europe, and the rest are in the U.S. About 70 people are in the U.S. but we are in 14 different states in the U.S. We have our employees in North Carolina, Virginia and Massachusetts, and even in Idaho and all these places. We have one colleague there based out of Bangalore, we will soon have a small team and centre in India in the early second half of this year.
Q Are you revenue positive or cash flow positive, close to being profitable?
A: Cashflow positive will be a very nice day, but this is still in the early stages. But the revenue potential of the product that we are building is enormous. Both the products I mentioned, the customer engagement product, and the other platform product that we launched last month, are around helping enterprises safely deploy AI in a responsible high‑performance way, in a way they can answer questions about what they can be comfortable about. So both these products are aimed at every single business in the world.
