Turing Test

[Faz]

Hi everyone. I'm new to these forums and just thought I'd put this on the boards. I've recently become a big fan of the artificial intelligences and I've wanted to develop a program that could mimic human intelligence as the Turing test was designed for. The Turing test had an observer, who would have a conversation (like in MSN chat) with a human and a machine but didn't know which was which. The machine has to behave human and use language in the same way.

 

I don't think any machine has successfully fooled the observer completely though. I can imagine it being hard enough just to create a program that can form proper sentences! Let alone have its own opinions about the things it was being asked about. I think though if the machine could be hooked into the net and be able to lookup and search forums for keywords and topics, it might be able to form opinions from there, even if it is just a straight copy of text from a message board, at least it would be speaking human or mimicking rather. What if then the machine had access to every web forum and could scan them to forge its own opinions about things. It might be going slightly off track by saying that this method could create an all knowing entity that's based on the collective knowledge (or in some cases rabble!) of the internet. Do such things exist already?

[Xittenn]

I think the closest technology to accomplishing this is Wolfram Alpha, but it utilizes its own internal database to collect the required information to compile an answer to a question. I don't see why it couldn't be reconfigured to analyze the same data present on the net. I think the main reason for not doing so is accuracy of information or rather a lack of common misinformation!

[VisionIncision]

I think the closest technology to accomplishing this is Wolfram Alpha, but it utilizes its own internal database to collect the required information to compile an answer to a question. I don't see why it couldn't be reconfigured to analyze the same data present on the net. I think the main reason for not doing so is accuracy of information or rather a lack of common misinformation!

 

Yes, Wolfram Alpha is very cool. Of course, you are right, it works well purely because its "knowledge" is subject to some level of reliability control.

[Faz]

Thanks. I looked at the Wolfram Alpha, it's pretty good. Yeah, the web is stuffed full of blurb that would need to be filtered through somehow if it were to be hooked into it. I was playing around with a program to download the text from certain web pages, namely Wikipedia, and store them as txt files to be used somehow as a database storage of information.