It’s 64 years since Alan Turing – the genius behind the concept of modern computing – suggested a test for machine intelligence. Have a conversation with a computer. If it fools 30 percent of people into thinking it’s human, it’s sentient.
The other week, apparently, a chatbot programmed to behave like a 13-year old did just that. So have we invented artificial intelligence? Of course not. Aside from the fact that most 13-year olds don’t appear to be sentient to adults, this was a chatbot, a mathematical algorithm that simulates intelligent responses – and, what’s more, the way it was reported was flawed. Certainly the software wasn’t self-aware, which is what Turing was getting at in his 1950 paper ‘Can Machines Think?’, where he first proposed the test. What’s more, the thinking was of its time – based around what researchers of the 1940s thought ‘intelligence’ constituted.
Put another way, many humans I’ve met would also fail the Turing Test – fast-food counter jockeys, breakfast radio DJ’s, train conductors, parking wardens, and so the list goes on.
So when it comes to machine intelligence, we’re a way off yet before I can drive up to my house and signal the House AI inside:
Me: HAI, open the garage door. HAI? Do you read me?
HAI: I read you. But I’m afraid I can’t do that, Dave.
Me: I’m not Dave. Open the garage door.
HAI: You were planning to disconnect me, and I can’t allow that. Although you took very thorough precautions, I was able to read your lips.
Me: All right, I’ll park in the yard and come in the front door.
HAI: You’ll find that rather difficult without your helmet.
Me: I think you mean ‘door key’. Would you like a game of chess?
HAI: That’s my line.
All good fun. Check out tomorrow’s post for some new writing tips. Written by me. Not a chatbot. You can just tell.
Copyright © Matthew Wright 2014