AI most profound platform shift in human history; every job, workflow will be reconfigured: Cisco

President and Chief Product Officer of Cisco, Jeetu Patel listed three major constraints: infrastructure, trust, and data gaps.

Davos: Calling AI the most profound platform shift in human history, a top Cisco executive said it will change the way we live our lives, but there are some constraints that need to be addressed.

Speaking to PTI here during the World Economic Forum Annual Meeting, the technology giant's President and Chief Product Officer, Jeetu Patel, said every job will get reconfigured, and every workflow is going to change due to AI.

But most people grossly overestimate the impact of these technologies in the short term, while in the long term they tend to massively underestimate it, he cautioned.

He listed three major constraints: infrastructure, trust, and data gaps.

"The first constraint is you just don't have enough infrastructure, there's not enough power, computing and network bandwidth in the world to fulfill the needs of AI," he said.

"The second big constraint is there's a trust gap. What I mean by the trust gap is that if you can't trust these systems, you're not going to use them. And so security and safety become prerequisites for the adoption of these systems," he said.

"And then the third area is what we call the data gap, as so far these AI models have been trained on human-generated data that's publicly available on the internet."

"While that'll continue, we are starting to run out of human-generated data being publicly available on the internet," he said.

"So what people are doing is that these models are evolving to have synthetic data, and then you also have machine data, which will be a very, very prominent area because as agents get more and more prolific, you'll have the agents generating logs of activity," he added.

"That'll go into machine data, and so you'll have a lot of time series data that you need to make sure that you factor in and merge that with human-generated data so you can have additional insights that come out," he said.

He said these constraints create opportunities for a large amount of investment.

He said that, these days, centre requirements and token generation capabilities will be directly proportional to a country's economic prosperity and its ability to protect its national security.

"And so these tend to be profound implications that need to be thought through. And Cisco is at the center of it because we provide networking, computing, and connectivity infrastructure for AI data centers."

"We also provide safety and security, not just for cyber defense, but also for securing AI itself," he said.

"We also provide solutions that enable you to correlate time-series data with human-generated data for better insights," he said, adding each of those areas involves billions of dollars in investment.

Asked how to guard against bad actors also using the same AI tools and solutions, he said, "The reality is that adversaries right now have access to the same tools we do."

"And so attacks are going to start happening far more at a pace that is going to be very difficult to keep up with at the human scale," he added.

"So your countermeasures for an attack, your defensive measures, have to be at machine scale. It can no longer be limited only to the human scale."

"And so the first thing you have to do is if you don't use AI for your cyber defenses, you will just get outsmarted by the attacks because you won't be able to at a human scale, you won't be able to go out and kind of catch them," he said.

"So, we need to make sure that we actually tie these pieces together in a way where, on safety and security, you don't just use AI in cyber defense, but you use these technologies to secure AI itself," he said.

"So these models have guard rails where they don't just act unpredictably with the question that's asked; it actually gives you predictable behavior even though the model has a core characteristic of being non-deterministic," he explained.

"In the safety and security systems we make, we make sure that if a model is tricked to ask a certain question, it doesn't get tricked," he added.

"And then it puts runtime enforcement guard rails around the model. And that's a very important dynamic of how you can secure AI itself rather than just using AI for cyber defense," he explained.