Google DeepMind Chief Scientist Claims AI Now Outperforms Humans in Most Tasks

The question of whether the new models will soon be able to outperform humans in the majority of jobs we perform now or achieve a level of Artificial General Intelligence (AGI) has been a hot topic in the AI community as artificial intelligence systems continue to advance.

Although he avoids using the term ‘artificial general intelligence’ (AGI), Jeff Dean, the chief scientist at Google DeepMind, thinks that, with few exceptions, current AI models may already have outperformed humans in the majority of daily jobs.

Dean claims that the existing models are “better than the average person at most tasks” that don’t involve physical exertion on an episode of the Moonshot Podcast. “You know, some of the models we have today are actually pretty reasonable at most things,” Dean said, “but most people are not that good at a random task if you ask them to do something they’ve never done before.”

What did Jeff Dean say about AI vs humans?

“How far away are the current models from making breakthroughs faster than humans?” the host then asked Dean. In response, Dean stated that AI models are currently on the verge of becoming perfect in a few areas and that he believes we will expand that range.

‘That’s a very different concept,’ Dean warned, ‘because most people will fail at many things; they’re not human experts in some areas.’ “There will be a lot of domains where automated search and computation actually can accelerate progress — scientific progress, engineering progress,” he stated. “All these things I think are going to be important for advancing what we as people can do over the next five, 10, 15, or 20 years,” he said.

The AGI Debate

However, Dean decided to avoid the AGI controversy, partly due to the lack of a precise definition for the term in the artificial intelligence community. “The difficulty of the problem varies by factors of a trillion, and many people have very different definitions of it, which is why I tend to avoid AGI conversations.”

More upbeat about AGI is Dean’s boss, Demis Hassabis, CEO of Google DeepMind, who stated in a recent interview with WIRED that the breakthrough would be made within the next five to ten years.