Will AI really outsmart humans?

Computers have beaten humans in chess, Go and, most recently, rock-paper-scissors. Could it be that artificial intelligence (AI) will surpass human intelligence in the near future?

This topic was taken up at the Rakuten Technology Conference 2016, where two experts shared their thoughts on the likelihood of AI ever matching human intelligence – and also on some new approaches that might make such an outcome a little less worrying.

Laurent Ach, Manager of Rakuten Institute of Technology Paris, reminded conference attendees that in March this year AlphaGo, a computer system developed by Google DeepMind, had defeated a highly respected human opponent in the popular East Asian strategy board game of Go. It was the first time a computer had ever beaten a human in the game without the aid of a handicap.

“No human player would have played those moves,” Ach said of AlphaGo’s play. “It seems that one after another, a computer will outperform the most skilled humans.”

Yet Ach, who began working with artificial neural networks in the 1990s, was quick to add that despite AI’s headline-grabbing achievements in limited tasks like Go, he still thought computers had little chance of matching humans in developing a sense of general intelligence.

To illustrate, he noted how chatbots have been around for 50 years, but still can’t carry on a convincing conversation. The problem is that they don’t actually understand what they’re talking about, which is linked to the issue of whether AI can replicate human understanding.

Ach contrasted philosopher John Searle’s take, which is that computers are merely simulating some part of human reasoning, with that of futurist Ray Kurzweil, who believes computers will achieve consciousness and quickly surpass human intelligence. Ach is confident Kurzweil’s prediction will come to pass – he’s just not as bullish as some on how soon that will be. “It’s not on the horizon,” he said.

But AI is also a deeply cultural phenomenon, something Rikkyo University lecturer Mary Reisel studies. Speaking during the same session as Ach, Reisel – a cyber anthropologist focused on investigating the psychology of online communities – emphasized the importance of ancient religious taboos against creating human-like intelligences, which account for some modern-day fears about AI.

“Humans have been the same for thousands of years,” Reisel said. “We have the same brains, we have the same fears, we have the instincts. We haven’t changed that much. It’s technology that has changed.”

She positioned the rise of AI in a religious and historical framework and painted its possible future as either a path to dystopia or one to a new human Renaissance. While she often feels pessimistic about AI’s growth, she says there’s also reason to be hopeful.

“The AIs that exist today are making ordinary users dumber instead of more intelligent,” Reisel said on the sidelines of the conference. “I would think of a simple app: instead of ‘Let me tell you what books you like,’ it would be ‘Let’s think together what you’d like to read now.’ AI should make the user think, choose or be creative.”

Read more posts from the Rakuten Technology Conference 2016 here.