Chevron icon It indicates an expandable section or menu, or sometimes previous / next navigation options. HOMEPAGE

A futurist who's right 85% of the time says machines will be conscious by 2025 — and it'll be 'the beginning of the end'

terminator genisys concept art seattle
Courtesy of Paramount Pictures

Google DeepMind's artificial intelligence AlphaGo made history when it won the complex game of Go against Lee Sedol, one of the greatest world players.

Advertisement

As Elon Musk pointed out at the time, experts in the field thought AI was a decade away from reaching that milestone. The momentous event showed that AI was gaining skills typically reserved for humans far faster than we expected.

And that very fact could be a problem, Ian Pearson, a futurist with an 85% accuracy track record, told Tech Insider.

"You could end up with superhuman machines going down that road," Pearson said. "Google's DeepMind isn't there yet, but really I'm sure they'll probably discover things along the way and, by 2025, it's possible their computer could be superhuman and could be conscious."

"That could be the beginning of the end, really," he added.

Advertisement
ex machina
A24

Pearson isn't the only one who thinks we're inching closer to seeing machines with human levels of consciousness. Ray Kurzweil, an AI expert for Google, wrote in his book "The Singularity is Near" that current efforts to reverse-engineer the brain will allow us to simulate it in computers by 2030.

The more computers can think like humans, the better they'll be at performing tasks. We often take for granted how easy it is for us to do basic tasks like, say, reach into a fridge and grab a beer. But asking a robot to do that same task is really difficult right now: it needs to understand what it's reaching for, how to get it without breaking or knocking anything over, and how to transport it safely.

If we want AI to do everything from buy us plane tickets through our phones to powering robots that help around the house, it needs to understand the world more like humans.

But there's also risks associated with AI evolving to the level of consciousness Pearson and Kurzweil are expecting within the next 14 years.

Advertisement

"Advanced AI could write its own codes and algorithms and take over other machines to make secondary AI, and people may not know where it is, nevermind how to switch it off," Pearson said.

AI with human levels of consciousness could also figure out how to work around any restrictions it was initially programmed with, he added.

"There's a certain amount of naivety [with the idea] that you can explain the task and it'll do so," Pearson said. "People underestimate the potential for a smart machine to choose not to follow the rules."

elon musk
Tech Insider

Tesla CEO Elon Musk has been particularly outspoken about the threats AI poses as it gets smarter.

Advertisement

He's funded a number of research projects to ensure AI doesn't turn evil, and he's not afraid of hyperbole on the subject: "With artificial intelligence we are summoning the demon," Musk said at MIT's Aeronautics and Astronautics department's Centennial Symposium in 2014.

In fact, Musk formed OpenAI, a non-profit research company, that's focused on advancing "digital intelligence in the way that is most likely to benefit humanity as a whole." An OpenAI researcher recently teamed up AI experts at Google to outline "Concrete problems in AI safety."

So it does seem some experts in the field are aware of the potential threats posed by advanced AI. Still, Pearson said he is skeptical of people who think it can be easily controlled.

"Everyone wants to get all the benefits without any of the problems, but if you look at any area of engineering or bioengineering you see things going wrong occasionally and that's in spite of people's best interests," he said. "It can't easily be constrained."

Advertisement
Close icon Two crossed lines that form an 'X'. It indicates a way to close an interaction, or dismiss a notification.

Jump to

  1. Main content
  2. Search
  3. Account