AI software takes on Turing

AI software takes on Turing

www.pcpro.co.uk

Artificial Intelligence software came close to beating the Turing Test for the first time over the weekend, scoring just 5% below the required pass rate.
The Turing Test, developed by English Mathematician Alan Turing, is designed to measure the ability of a computer program to mimic human conversation.

In the test, human integrators communicate via a terminal with either a human or a computer simulation. If a program can convince 30% of judges that it is human, then it passes the Turing Test. However, since it was devised in the 1950s no machine has ever passed.

"Although the machines aren't yet good enough to fool all of the people all of the time, they are certainly at the stage of fooling some of the people some of the time," says the organiser of the Turing Test, Professor Kevin Warwick from the University of Reading's School of Systems Engineering.

"Today's results actually show a more complex story than a straight pass or fail by one machine. Where the machines were identified correctly by the human interrogators as machines, the conversational abilities of each machine was scored at 80 and 90%.

"This demonstrates how close machines are getting to reaching the milestone of communicating with us in a way in which we are comfortable. That eventual day will herald a new phase in our relationship with machines, bringing closer the time in which robots start to play an active role in our daily lives"

(visit the link for the full news article)

Think this is pretty significant, if computers are becoming proficient at conversing then surely the though process behind it must be similar to that of a small child?

I would imagine the first stage of true AI must be to acquire the logic or thought process. Once complex conversation is acquired then it's a first step for a computer to be able to learn for itself from external inputs?

Maybe real AI isn't too far around the corner then?

www.pcpro.co.uk
(visit the link for the full news article)

Its not AI though is it.

All the computer is doing is responding to input based on a pool of previously associated words and phrases. That is not intelligence, thats just a complex challenge / response system. There is no deliberation or opinion or elaboration of abstract or imaginative concepts, just the illusion that there is.

yes, i agree that it isn't really AI

i had fun with it for about 5 minutes playing word association but then kept reminding it that it was giving vague, predetermined responses to my questions.

oh btw, it will get angry with you if you ask about its trash can


it did with me anyway so that was fun lol



***edited to finish typing out my post. my tired mind failed to catch the fact that i'd missed a chunk of what i'd intended to type

[edit on 13-10-2008 by Barathrum]

reply to post by Barathrum


Haha nice one! Mac's can also be pretty huffy when you ask them to eject a floppy!

It must take some level of AI though to recognise and respond to conversations surely?

i'm not sure it would really be considered AI

granted, it can give responses to questions but when trying to converse with it, it becomes apparent that it doesn't have true understanding of what you are asking. also, it doesn't seem able to learn from conversation or to really even have a true conversation, as it tends to reply in such a way as to attempt to lead the conversation.

it is rather fun though and i think if they keep developing it that in a few more years it will be quite impressive.

Well it depends on your definition of intelligence i suppose.
But essentially all it is doing is responding to your input, like a calculator, would you class that as intelligent? Its clever, but at the end of the day its just a straight forward logic system, and intelligence is alot more than that.

reply to post by Man_Versus_AntiMan


Yeah you've got a good point there I suppose.
Would be interesting to see the next generation of role playing games with responsive verbal inputting!

there is a problem in the definition of what is AI;

if you think at "ghost in the shell" (mustwatch) or steven spielberg, then ok but i'm sorry because you are wrong.

AI is just a way to program machine, introducing semantics and self-learning to those machines; a new and revolutionary approach to programming (that is there for sometime now, 20years maybe at least)

an example of AI technology is neuronal network that are used for example to predict traffic on a road network and therefore avoid them
another are knowledge base system (different that database)

this machine is AI.
review your definition of the term please.


[edit on 13-10-2008 by ::.mika.::]

reply to post by ::.mika.::


Its not the definition of Artificial Intelligence that needs reviewing, it suggests exactly what it says on the tin. I would say the use of the term intelligence in this context is inaccurate and misleading.

If making preprogrammed judgements based on external inputs is intelligence, then why do we not class an excel spreadsheet, calculator or computer as intelligent?

If responsive traffic systems are really classed as AI then it is the fields use of the term 'intelligence' that needs reviewing, surely.

it's not the final product to be classified as ai or not ai, it is just the way it is being built.

the outcome doesn't come from a preprogrammed process, the process is created by the system providing that it was given enough examples (or experiences) upfront (input-output).
we are not in a case where : if input then output. but in a case where the machine looks the situation and depending on its own knowledge base ( experience) will create the output on its own.

excel, calculators are in the "if then" logic and it is not AI,

this machine that seems articulated on a knowledge base that it self-fills as its experience grow is AI.

but I agree totally that the terminology chosen when this new generation of programming was developped is totally debatable.

(just like those crazy bankster that called themselves economist and chose to call their model "liberal"...)

[edit on 13-10-2008 by ::.mika.::]

[edit on 13-10-2008 by ::.mika.::]

[edit on 13-10-2008 by ::.mika.::]

actually far less than the banksters because this way of creating the process to generate the output as those machines do is very close to our ways of reacting as "intelligent entities".

and they can be conditioned as well with false knowledge that makes them # up.

Interesting stuff,

I remember hearing about the Turing test when i was a kid, it used to be quite a scary thought that machines might one day reach that level. Now we're almost there it feels like a bit of a skank.

Im sure Turing himself envisaged the machine that would pass the test would display actual intelligence; judement, reasoning ability, even self awareness. After all thats the whole purpose of the thing.

Now tho its obvious the first machine to achieve this will only be mimicing intelligence, like a parrot that can chat all day with zero understanding of its words.

Kind of renders the whole test insignificant imo.

[edit on 13-10-2008 by Man_Versus_AntiMan]

I worry that we are looking for machines to show intelligence in a conversation, but that requires many strands of intelligence. I think a lot of people are looking at machines for fighter capacity being able to take up to 16g's while pilots can take 6. I wonder if our whole air forces were taken out by these critters, we would consider them as intelligent whatever the judges say....

It was interesting but completely unable to process my responses. It is a long way from being confuseable with any human I would normally associate with. It had a schizophrenic flair to it and really just gave the impression of being nothing more than a fancy magic 8-ball.

Originally posted by Man_Versus_AntiMan
display actual intelligence; judement, reasoning ability, even self awareness.
[edit on 13-10-2008 by Man_Versus_AntiMan]

watch ghost in the shell, 1 and 2

we really don't want this to happen, ever.

as rabelais said "science without consciousness is ruins of the soul".

some application of AI programming could be/are already however very interesting, efficient, sustainable and useful.
but the researchs in those fields, just like in bio-genetics, should require a set of really really strong boundaries, limitations, or sooner or later, as technologies grows without limits, it will become totally out of any control. the very essence of the concept of AI programming making it certain.