Global IT service provider Cognizant Thursday unveiled the launch of its Advanced AI Lab.
The new lab, based in San Francisco, is not Cognizant’s first AI-focused lab, said Babak Hodjat, CTO of AI at Teaneck, N.J.-based Cognizant.
“We do have AI Labs, and we have AI Innovation Centers,” Hodjat told CRN. “This is an Advanced AI Lab, which is focused on core AI research and breakthroughs. So in that respect, is it’s quite different than our other AI Labs.”
[Related: ServiceNow President Desai: Generative AI Technology Strong Out Of The Gate]
Cognizant set up its new Advanced AI Lab in downtown San Francisco overlooking the Salesforce Park stadium because the city has become a hotbed of AI activity, Hodjat said.ADVERTISEMENT
“San Francisco is turning into a center of gravity for AI,” he said. “There’s a lot of AI startups and smaller companies. And for some large companies, their AI centers are here in the city or in the [nearby] peninsula. So it doesn’t make sense for us to not be situated here.”
When Hodjat described the new Advanced AI Lab as focused on “core” AI technology, he referred to a focus on developing new use cases and ways to take advantage of AI.
“Typically, AI companies look at industrializing AI and creating use cases, which is part of our mandate,” he said. “But we’re going beyond that. We’re looking at actually expanding the state of the art in AI itself, coming up with new algorithms and technologies that would be useful for our clients. And in that form, we will be a few years ahead of the game as far as technologies that might actually get to our clients. But it will be groundbreaking and will allow us to be differentiated from an algorithm and technology perspective, not just from a perspective of being able to create the solutions, which we are already pretty good at.”
The primary use of AI today is centered around improvements in productivity, creating better natural language-driven user interfaces, summarizing data or helping with coding.
We want to expand that,” he said. “We’re looking at every piece of technology in an organization and ultimately even AI and modeling is at the service of making some decisions. So we view use cases of AI from the perspective of the ultimate use case, which is making decisions. When you look at it from that perspective, a number of components and algorithms become very important.”
One very important element is trust, Hodjat said.
“When can we trust the AI?” he said. “And when should we trust it less? In other words, when can we defer to the AI or take its suggestion for a decision more seriously? And when should we actually have humans take a look? That’s a whole field of uncertainty and uncertainty modeling. We do have some patents on it, which basically gives us a sense of for every single prediction that the AI is making to look at how familiar or unfamiliar that decision and the context of that decision is. And based on that, it will tell us ‘You should be this much certain or this much uncertain.’”
Based on that information, businesses can set a threshold and a measurement based on that certainty, Hodjat said. “And if it’s below that threshold, you can actually have humans take a look.”
Hodjat said advanced AI is also aimed at what he called “explainability.”
“We live in a world where the state of the art in AI is simply not explainable,” he said. “These are neural networks, generative AI, large language models. These are all deep-learning neural networks that are inherently opaque. But when we want to make decisions, we want to be able to ask the question, ‘Why?’ Like, ‘Why do you recommend this? What did you look at? Why do you think this is the right decision versus this other decision?’ And so in spite of the fact that, through the workflow of coming up with that decision we are using opaque models, at that last step we’re able to actually make it explainable. That’s a whole field onto itself.”
Most decisions are aimed at more than one outcome, such as during the pandemic when people were thinking of reducing the number of patient cases while at the same time minimizing the economic disruption to people’s lives, Hodjat said.
“And when you think about it, they’re kind of opposing goals,” he said. “At one extreme, you can shut down the economy and save lives, but then lose lives in the longer term. On the other extreme, you can just leave everything wide open and the economy’s rolling, but you’re losing lives. Often our decisions are what we call multi-objective, where we’re optimizing for more than one outcome. That’s a whole other very, very important field. You know, we want to increase revenues while reducing costs, while being responsible, while being ethical. These are all goals, and they should all factor into our decision-making.”
Cognizant’s AI platform currently allows the company to build use cases tailored to clients within minutes, Hodjat said. It uses generative AI to help scope the use case, generate synthetic data that resembles the data that the client will use, and puts together an entire workflow, all within minutes, he said.
“I actually create them in front of clients,” he said. “Our associates are trained to build them in front of clients. That’s our vision. Our vision is that we show the client the art of the possible and how they can disrupt their workflows through AI-based decision- making.”
Cognizant’s new Advanced AI Lab is open and has 18 people, Hodjat said. It expects to have about 30 total by year-end. As part of that growth, it expects to triple the number of Ph.D.s on staff and bring in some other functions, including interns, as well, he said.
Hodjat declined to discuss the monetary investment Cognizant made in its Advanced AI Lab. However, he said the investment was part of a larger three-year, $1 billion investment Cognizant previously said it is making.