Wednesday, October 17, 2007

British Telecom's futurist predicts end of the world in about 10 years

Ian Pearson tries to predict the future for British Telecom. I think I've previously written about his 2006 Technology Timeline.

Now he's being interviewed by Computerworld about the develop of sentient machines.

Computerworld - BT Futurist: AI entities will win Nobel prizes by 2020

...We will probably make conscious machines sometime between 2015 and 2020...

... I think that we still should expect a conscious computer smarter than people by 2020. I still see no reason why that it is not going to happen in that time frame....

... they will get very, very clever. It's kind of like a hamster trying to understand a human being. They can't simply understand the problem. How could they possibly think in the same way? It's like as if a human being is compared with an alien intelligence, which is hundreds of millions of times smarter. We don't have the right capabilities to start thinking in the same way. So, we put machines winning Nobel Prizes in our technology timeline, because we got good reasons to do that...

The scenario is very familiar to anyone who's done their essential reading. Certainly I've written about it enough. Once machines get to hamster level, much less consciousness, it's basically game over. I'm not entirely confident humanity will vanish immediately; that might depend on the AI's sense of humor.

What's novel is the date -- that's the earliest prediction I've read from anyone gainfully employed. Most of predictions are out around 2040, where I have a decent chance of being safely oblivious (whether dead or alive). In ten years I might be still standing.

I think he's wrong. Actually, if I were the praying type, I'd pray he's wrong. I'm more inclined to 2050 myself, which makes it my kids' problem. Once you move it out to 2050 there's a decent chance of prior civilizational collapse anyway, which could give Homo sapiens a bit of a longer run.

BTW, wouldn't a TV show about a top secret spy organization that goes around the world messing up AI projects be a lot of fun?

No comments: