Comments on: Chatbots try to impersonate humans – and fail

By: David King

David King — Fri, 17 Oct 2008 06:32:09 +0000

I'm in computer support. I've worked with personal computers since before Windows and DOS. (This just to establish my credentials). About 10 years ago I was at a home show and came across a booth selling the Encyclopedia Britanica (20 printed volumes). I told the salesman I would wait for it to come out on a disk. The salesman said "Not in our lifetime!" I said "5 years". You can now buy the Encyclopedia Britanica on a DVD. This is leading up to my opinion that a computer that can pass the Turing test will happen in less than 10 years, likely less than 5 years.

David K.

By: Chip

Chip — Mon, 13 Oct 2008 14:10:27 +0000

I had to go try Elbot out — I was disappointed that I was so easily able to stump him. Just ask "which foot do you have forward right now?" He'll try to change the subject every time. If you do it enough times, though, he replies, "I'm afraid you're in an endless loop. Maybe you should re-boot yourself."

By: Chip

Chip — Mon, 13 Oct 2008 13:47:14 +0000

But to pass, a computer does not have to mimic the human brain perfectly — it just needs to be able to fool one. We humans are pretty easily duped. For millenia we thought rocks and trees were possessed of humanesque intelligence. We find it comforting to project anthropomorphisms onto non-human agents, so we're likely to help the Turing test winner along, whether we mean to or not.

By: Nathan Zeldes

Nathan Zeldes — Mon, 13 Oct 2008 12:58:49 +0000

Hey, Chip, I'll be delighted to live to be proven wrong… but I've produced Silicon chips (I work at Intel) and I've studied some Neuroscience in my time… the human brain has some 100 billion neurons, with up to 10,000 synaptic connections (inputs) EACH, and all working in parallel. Current Microprocessor chips have 2-4 parallel cores, and around 100 million transistors. Long way to go…

And of course the Turing Test is very aggressive – it suffices to trip the computer up on one single question…

By: Chip

Chip — Mon, 13 Oct 2008 11:14:29 +0000

I wouldn't be so sure. Back in the '50s some of the smartest scientists and science fiction writers thought that we wouldn't get to the moon for generations — but it was only a decade away. Sometimes a single big push or breakthrough will make all the difference.

By: Nathan Zeldes

Nathan Zeldes — Mon, 13 Oct 2008 09:01:14 +0000

I wouldn't worry… a computer passing a full Turing Test will happen long, long way in the future, if at all.

And check my post on the reverse problem here…