Passing the turing test is the holy grail of artificial intelligence. A lot of recent books claim the computing
power needed for such an AI is rapidly approaching while some still claim that it is impossible without a quantum mechanical soul. What do you think?

I believe it will be possible to pass the test within five years. If we concentrate on building a textual-based intelligence (no worries about motors, visual detection, etc.), I think this might be doable.

My worry is that a lot of effort may go into configuring a machine to uphold the illusion of intelligence and not the actual creation of intelligence. I see the advancement of AI only happening after we give up the idea that there is only one brand of intelligence (namely the human brand). We think very highly of our own intelligence and communication skills and I am sure that some day machines will also take <i>pride</i> in their intelligence and ability to communication. To test a machines intelligence based on its ability to mimic our perception of intelligence is unfair. It is an apples to oranges comparison and if the machines where testing us this way, how would we fare?

You know, people were saying that when turing first came up with the test. Textual-based AI has been failing to deliver for >30 years now (precisely bcause symbolic methods by themselves don't work) and anyway the biggest problem in AI now is in deciding just how symbols are grounded in the world eg when you say 'I saw ' or 'I ran'... the meaning is highly integrated with the activities of your visual and motor systems respectively. Sure we have a 'language system' which can very easily deal with textual language, but the apparent order and ease of comprehension in that symbolic system are due to the constraints placed on it by our sensory-motor system.

And anyway so what if you pass the turing test? If the AI that does it isn't really intelligent then all you've got is a tool that can trick people real well --- trust me this could be done without the AI really being intelligent, it would just require the programmer to recognise the kinds of thought patterns that HBs all too readily fall into, and maintain with the same suicidal vigor with which a lemming maintains its stride even though its already reached the edge of the cliff. Point being that if it relies on the intelligence of the programmer, then the AI itself is not really intelligent --- thus not very interesting at all.

A few other people still ask whether the Turing test is valid. Of course its not. Humans have the level and type of intelligence they do for a specific reason. The problems we have had to evolve around has shaped our mental level and it is absurd and even insulting to try to replicate it in a machine. Computers will (or rather do) have their own goals and methods of solving problems. I have fifty zillion sperm that could create a machine to pass the Turing test. Surely we should try for something more useful and better.

INMO I think within 10 years it will happen.

ots is absolutly right, 10 years hands down =)

Of course 10 years, but I think we need to adopt a different strategy, not so complicated. Artificial Life needs only that one spark to come into being. After that the turing test will be kindergarden for bots of all types.