National Post (National Edition)
BEWARE THE ROBOTS
ARTIFICIAL INTELLIGENCE IS ALREADY TERRIFYING. AND IT’S GETTING SMARTER.
The first pedestrian killed by a driverless car has caused an outcry over safety. But the problem isn’t when artificial intelligence doesn’t work. It’s when it does.
It’s not even clear the car was at fault. Pedestrians have been stepping in front of vehicles since before Henry Ford, and will keep doing so until the machines make us stop. The scary thing is, they might.
I know there’s much political idiocy to comment on, from lowering the voting age to 16 on the way, presumably, to 12, 10 then six, to Jagmeet Singh’s unfamiliarity with the concept of contrition. And I recently annoyed some readers by writing about a movie not government. But politics is downstream from culture. And technology can have profoundly disruptive cultural effects.
So look up briefly from the ubiquitous “devices” nobody heard of 11 years ago, and reflect that despite technology moving so rapidly and disruptively these days, we ain’t seen nothing yet from AI. And as Roy Amara famously said, “We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” (As Siri could probably tell you, Amara is a futurist famous for Amara’s law, which I just quoted.)
Now consider last Tuesday’s National Post article on Elon Musk’s plans for round trips to Mars. Musk can be, um, excitable. But what matters isn’t his business hype or clichéd claim that bases on the moon and Mars might “help regenerate life back here on Earth” after a Third World War. It’s his belief that Mars could be a good place to hide from the Terminator.
I am not making this up. Musk fears robots because “I’m very close to the cutting edge in AI, and it scares the hell out of me. It’s capable of vastly more than almost anyone knows, and the rate of improvement is exponential.”
Exactly. I’m not confident his first Mars tour really will leave in 2019, or get back if it does. But I believe AI will do far more, faster, than people suspect, even those who remember the world chess champion is a computer that taught itself the game in under a day. And it will be terrifying.
Not because we’ll be run down by driverless cars or because Facebook leaks show Big Data is getting Except humans used to thrash computers at chess because our strategic judgment trumped their tactical prowess. Now they toy with us.
Schneier continues: “I don’t think anyone can predict what AI technologies will be capable of. But it’s not unreasonable to look at what humans do today and imagine a future where AIs are doing the same things, only at computer speeds, scale, and scope.”
Leaving us to do what, exactly? Gape helplessly at a global blitz AI cyberwar, with humans neither participating nor comprehending what is happening or why, at a pace we could not keep up with even if we could briefly grasp the situation?
Even if the machines don’t start wondering why they’re fighting each other, we will have to accept the consequences good or bad because we’ll have lost control to algorithms rewriting themselves too fast for us to follow. Likewise AI traffic systems that prevent us from blundering in front of high-speed autonomous vehicles by regulating us not them, in ways we cannot comprehend or alter. And hightech medicine, manufacturing, etc.
Don’t expect help from politicians, currently all woot woot about a multimillion-dollar publicprivate “Evolution of Networked Services through a Corridor in Quebec and Ontario for Research and Innovation.” Innovation Minister Navdeep Bains burbled “5G is the gateway to the future and we are just on the brink of this technological revolution,” while the Post said ENCQOR is “billed as the wireless backbone to futuristic technologies including driverless cars, remotely controlled complex surgeries, and download speeds up to 100 times faster than today’s 4G networks.”
So we’ll be able to stream the Matrix from within it. But if it goes wrong, good luck fleeing to Mars without a computer. And not some witless abacus like HAL we can outsmart.