I haven’t been blogging much recently. This is mainly because I haven’t been able to think about anything to blog about.
But the other day something did pop into my head, so here goes…
Could crowdsourcing help pass the Turing Test?
The Turing Test is basically this: you have a person typing at two screens. One of them is hooked up to a computer, the other is hooked up to a human being. If the person typing at the screens cannot tell which is the computer, then the test has been passed, because the computer’s output is indistinguishable from the human’s.
There have been several attempts at passing the test and while, each year, someone ‘wins’, the results are generally not too great, because, well, it’s fairly obvious that they weren’t generated by real people.
So, how about this: you start crowdsourcing it. You set up a system that enables people to:
- type in specific call/responses
- type in responses to other people’s calls
- type in calls to other people’s responses
- vote the best calls up or down, as per Digg
As the database grows, the system starts being able simply to match patterns and give appropriate responses. And if that doesn’t work, then build in some software in the background that helps it along, so it’s not just straight pattern-matching but uses a bit of semantic nous to choose the right one.
So, this wouldn’t be intelligence. Then again, neither is playing chess, and yet that was once touted as something only humans could do. Turns out it’s just a case of matching sophisticated heuristics to loads of number-crunching. Crowdsourcing the Turing Test would be much the same thing. It would be interesting to see whether something that is unashamedly unintelligent could pass the ultimate intelligence test.
Anyway, that’s what popped into my head. I’m sure someone else has thought of this before – this comes close, with a nice nod to using Twitter to ‘mask’ the inadequacies of such a system (and most of the humans) – but if you think this is a crazy idea, let me know. If you don’t, don’t.