When will a machine pass the "Turing test"?

[ybbor]

The real test is if a AI can join CFC and get to 1,000 posts without anyone question it's humanity.

they already did it's...nah...it's too easy.

I think it will happen. Not in the forseeable future though, voted 2050 or later

 

[Winner]

Well, given the current development, I guess between 2020 and 2030.

What will happen next? Check the Battlestar Galactica 2004 or Terminator 3 for further info

 

[farting bob]

just because it can be done, NEVER means it should be done.

A large reason of why technology has advanced at such an incredible rate in the last 25 or so years is because people are preapred to push the boundaries, and this is just the next step.
I voted before 2020. At the exponential rate of advancment, it wouldnt surprise me if its within the next 10 years.

 

[newfangle]

In 3 years when a friend of mine finishes his Ph.D.

 

[stormbind]

Already been done. Computers can hold conversation for 10 minutes and fool a large number of people.

My own computer held a conversation with a Greek programmer on ICQ. The Greek guy thought I was drunk but persisted to hold the conversation for 30 minutes before accusing "me" of being impossible.

I kept some of the message history somewhere for laughs

I am sure my PC would have performed similarly in a 3-way conversation.

Anyway, the Turing test does not specify for how long the computer must remain anonymous, so I say it has already been done.

 

[Sparta]

I am sure my PC would have performed similarly in a 3-way conversation.

What a glorious day it will be when my PC can handle conference calls for me (preferably without sounding drunk, but I'll take what I can get).

 

[Oerdin]

It's 30 years out atleast likely even longer.

 

[Giio]

has anyone tryed talking to alice?
If so have you said exactly what Alice 1 says in the conversation between the two?

 

[aaglo]

With quantum computers it might become possible - I vote for 2050 or later.

 

[Giio]

What are quantam computers? Ive never heard of them before.

 

[aaglo]

quantum computers are still a bit on the science fiction side, but people are working on building those. They are computers capable of making almost infinite number of calculations simultaneuosly, because of their new way of handling binary numbers (0's and 1's).

You see, in current computer, each bit must be either 0 or 1. In quantum computer, each bit can be 0 and 1 simultaneously. How is that possible, I don't know - I haven't looked into it very deeply.

 

[Eetu Pellonpää]

First I thought this was about "Turing's machines". Aren't those robots, that can soveirgly make a copy of themselves? I recall there was an article in a newspaper, that some of such bots were made succesfully. Well...

By the way, the poll measures when the machines will pass the turning test...

 

[Aphex_Twin]

The Turing Test has no standards, it's impricise. It's basicly an "I'll know it when I see it" thing. With science precision is required or else it has no meaning.

Not necessarily. While a bot might be able to fool one human, it can't fool all. If there are enough people being "fooled" by such a machine, or perhaps enough people with expertise - we could conclude that until further notice that machine is indeed thinking. If it walks like a duck and quacks like a duck...

As for machines passing Turing tests, they have been since the early days of AI.

 

[farting bob]

quantum computers are still a bit on the science fiction side, but people are working on building those. They are computers capable of making almost infinite number of calculations simultaneuosly, because of their new way of handling binary numbers (0's and 1's).

You see, in current computer, each bit must be either 0 or 1. In quantum computer, each bit can be 0 and 1 simultaneously. How is that possible, I don't know - I haven't looked into it very deeply.

Why qould Quantum computers be needed? It will provide much faster processing i presume (havent looked into them in great detail), but for AI conversation, processing isnt the most important thing. It wouldnt take huge amount of processing to work, so unless quantum computing is fundamentally different, not just faster, theres no need for them when taking the turing test.

 

[Aphex_Twin]

First I thought this was about "Turing's machines". Aren't those robots, that can soveirgly make a copy of themselves? I recall there was an article in a newspaper, that some of such bots were made succesfully. Well...

By the way, the poll measures when the machines will pass the turning test...

A Turing machine is everything that has an internal set of states, that can modify it's environment and change it's internal states based on the environment. A "universal" Turing machine is a Turing machine that, given infinite memory, can mimick the behaviour of any other Turing machine. Your personal computer is a Universal Turing Machine.

 

[wit>trope]

Turing test actually has to do with sex. The wikipedia link should talk about it

 

[puglover]

Never. You cannot programm creative thinking into a machine.

I agree. You can probably put some of the inventor's creativity into the machine, but the machine will never be able to come up with creativity of its own.

 

[Aphex_Twin]

I agree. You can probably put some of the inventor's creativity into the machine, but the machine will never be able to come up with creativity of its own.

What makes you so sure?

 

[Rik Meleet]

Already been done. Computers can hold conversation for 10 minutes and fool a large number of people.

My own computer held a conversation with a Greek programmer on ICQ. The Greek guy thought I was drunk but persisted to hold the conversation for 30 minutes before accusing "me" of being impossible.

I kept some of the message history somewhere for laughs

I am sure my PC would have performed similarly in a 3-way conversation.

Anyway, the Turing test does not specify for how long the computer must remain anonymous, so I say it has already been done.

That's not the Turing test. For the Turing test you need to have 2 entities (1 human and 1 computer) and a human panel. The human panel asks questions to the 2 entities. The human's goal is to be identified as the human; the computer's goal is to be incorrectly identified as the human.

If the panel asks questions like: "how long did the pink hippopatomus fly under the Missisippi river?" - the PC succeeding the TT has to realise that hippo's aren't pink (or any other non-hippo color) as well as realise a hippo doesn't fly (without help of a flying device) as well as it is highly difficult to fly under a river.
When the panel makes up questions the human can easily identify as nonsense; the computer -who cannot think creatively- will answer these questions with "3 hours" (or a different timeperiod)

If you make your PC to answer creatively, the panel will slip in a normal question, like: "What was the cause of death of Ronald Reagen?" and the entity answering in a fake-creative way is not the human (remember the goal of the human is to be correctly identified as the human).

 

[Masquerouge]

First I think there is much more to intelligence than speech. So Turing will not get us anywhere.

On the other hand, I'm pretty much convinced that there is nothing more to our intelligence that chemical and electrical reactions, probably coupled with topological mathematics. So there's nothing in theory preventing us for making an artificial brain... as soon as we have figured how the real one works.
Or, why not try a completely different approach ? Who knows...