Hype vs reality: Is the tech industry on the cusp of another ‘AI winter’?

End-users should be wary of getting too caught up in the industry hype surrounding artificial intelligence (AI), because history suggests it could be relatively short-lived, warns Stanford professor and tech investor David Cheriton.

During a panel session at the NetEvents Global IT Summit in San Jose, Cheriton said the technology had been through a “bunch of hype cycles” since as far back as the 1960s that have resulted in the benefits of AI being oversold, inflated and, in turn, the market losing interest.

This is based on Cheriton’s observations and experiences of working at Stanford for the past 38 years, as well as his time in the tech industry as co-founder of Arista Networks and an early investor in Google and VMware.  

“Back in the 1960s, the view was we’re just on the edge of producing machine intelligence,” he said. “It kind of died out, and then didn’t go anyplace. In the late 1980s, the view was expert systems were going to take over the universe. There were some very impressive projects done at Stanford there.

“There was great concern that the Japanese with their fifth-generation computer projects… were going to just dramatically leapfrog the US in technology.”

Instead, what followed was the onset of an “AI winter”, where hype of the technology gave way to reality, ushering in a period of reduced funding and interest.

“When people have asked me over the last 35 years in Stanford what I thought of AI, I say well, it’s a very promising technology,” said Cheriton. “It’s been promising ever since I encountered it and continues to promise, but I think it suffers from being over-promising.  

“You look at the excitement over AI, you look at all these articles and say well, people are going to be thrown out of jobs and AI machines will take over the universe, and so on. I’ve been there, seen that and then what follows is an AI winter. There’s a certain basis for scepticism just on experience.”

Expanding on this point, fellow panel member Ravi Chandrasekaran, senior vice-president of the enterprise networking business at network tech giant Cisco, spoke of the need for end-users to manage their expectations about the range of use cases that AI can be applied to.

And in that context, it would be better to view the technology as something that can augment or assist with processes that humans may find harder to do on their own, and make them more efficient, said Chandrasekaran.

“We view these as tools that assist in certain things which we will not otherwise be able to do,” he said. “For example, if you look at encrypted traffic, you need some assist. You can’t easily find out what is inside [on your own].

“We look at that, and I think… the reality will set in [for end-users] where the reality [of AI] is somewhere in between. It is not nothing, but it is not the glory that people are talking about. We are starting to see some good results of that.”

The amplification benefits that AI can bring to the IT work that humans are responsible for within organisations was one area called out by Chandrasekaran during the panel as sign of the good that the technology can do.

Although a lot of the reporting on AI focuses on how its proliferation within enterprises could lead to job cuts, the converse is often true, he said.

“When we [Cisco] look at any IT organisation, they are growing,” he said. “They are hiring hundreds of people to run the network, or the digitisation that’s happening. What we see is that the [AI] tooling is basically to free them up from dealing with the complexity that comes along, so that they can actually get their job done.

“We look at all this automation, and… the idea is to free people so that they don’t become completely buried with the burden that’s coming along with the number of devices coming on board.”

CIOs are seeing the benefit too, as they know their IT department is liable to struggle without a little outside assistance, said Chandrasekaran.

“When Cisco talks to any of the CIOs or IT organisations, they’re so happy providing the tooling because they thought that they cannot keep up. They cannot run the infrastructure, it’s hard. They’re getting more people.

“Customers need this help. They understand these are the tools that you need to run the infrastructure. It’s not trying to replace people. It’s not trying to reduce the people. It is freeing them up.”