Search Episodes
Listen, Share, & Support
Listen to the latest episode
Subscribe via iTunes
Subscribe via RSS
Become a fan
Follow on Twitter

Support Us:

Please consider making a donation to help make this podcast possible. Any contribution, great or small, helps tremendously!

 
Subscribe to E-Mail Updates

Related Readings
  • Answers for Aristotle: How Science and Philosophy Can Lead Us to A More Meaningful Life
    Answers for Aristotle: How Science and Philosophy Can Lead Us to A More Meaningful Life
    by Massimo Pigliucci
  • Nonsense on Stilts: How to Tell Science from Bunk
    Nonsense on Stilts: How to Tell Science from Bunk
    by Massimo Pigliucci
  • Denying Evolution: Creationism, Scientism, and the Nature of Science
    Denying Evolution: Creationism, Scientism, and the Nature of Science
    by Massimo Pigliucci
Saturday
Jul262014

RS113 - The Turing Test

Release date: July 27, 2014

Did you know that an artificial intelligence named "Eugene Goostman" recently passed the Turing Test, our gold standard criterion for whether an AI is conscious? At least, that's what many media outlets breathlessly reported. In this episode of Rationally Speaking, Julia and Massimo take a critical look at Eugene, and at the Turing test in general as a standard for consciousness. In the process they debate what it would mean for an AI to be conscious, and how we could ever tell.

Massimo's pick: "Soccer and Philosophy: Beautiful Thoughts on the Beautiful Game "

Julia's pick: "Decisive: How to Make Better Choices in Life and Work"

References (24)

References allow you to track sources for this article, as well as articles that were written in response to this article.

Reader Comments (69)

I never heard of this Turing contest at the University of Reading where only 30% of the judges have to be convinced after five minutes.
I know the Loebner Prize, where half of the judges have to be convinced after 25 minutes for a silver medal, and there's a final audio visual input stage for a gold medal.

Funny that 13-year-old Ukrainian boys sound like robots. To make Goostman even more believable, make him an autistic troll.
Actually, many of Goostman's developers are Russians, which may have inspired his character, and they developed it in their spare time. It ain't IBM's Watson, which won Jeopardy.

The Goostman chatbot can only answer the most trivial questions by recognizing keywords like "weather" and "favorite," and tries to change the subject and ask questions itself.
But if a computer could really hold a conversation and answer non-trivial questions, that would be a big deal. It could replace all the Indian customer service representatives, who often robotically look up keywords and read off scripted answers themselves.
And if it could step in for Massimo or Julia and have an intelligent conversation on Rationally Speaking, then that would be human enough for me.

I wouldn't want my robot to be conscious though, so I can turn it off without being a murderer. It can be programmed to say "Ouch!" when kicked, but it doesn't need to actually feel pain. I'd rather have an intelligent robot that doesn't feel pain than a dumb but sentient robot that does.

July 28, 2014 | Unregistered CommenterMax

The funny thing is that the judges were having a parallel "control" conversation with a human, so those judges who thought Goostman was a human must've had a very robotic human whom they mistook for a bot. Maybe those humans were planted by Goostman's team ;-)

July 28, 2014 | Unregistered CommenterMax

It would be an interesting coincidence that the substrate necessary to construct biological organizations is the only substrate that could support consciousness, when the latter doesn't appear to be needed for survival.

Airplanes, in fact, can be constructed out of many different materials, even ones that would at first appear to be unsuitable, such as paper.

Julie's point that maybe many things have something like consciousness, but different, came to me from this video short that you've probably seen:

https://www.youtube.com/watch?v=8IjHxLm2kYw

July 29, 2014 | Unregistered CommenterGreg Esres

Massimo and Julia:

Regarding the issue of the nature of consciousness and its connection with the brain, please check out this Radiolab segment:

http://www.radiolab.org/story/anesthesia/

Somewhere in there the anaesthesiologist (a professor of anaesthesia) gives his own theory about the connection between the brain and consciousness. Now, whatever you may think right off the bat of an anaesthesiologist proposing his theory of consciousness, trust me, just give it a listen. I certainly don't think we'd want to say that he's solved it, but he might be onto something. In the very least, it's quite fascinating.

July 29, 2014 | Unregistered CommenterCheryl E. Fitzgerald

Seems more than a coincidence that Massimo, a former biologist, thinks that consciousness has to be biological. What's so special about the brain that can't be reproduced with an artificial neural network?
For that matter, why must life be biological? Robots can already respond to stimuli and power themselves with solar cells. If they could just build copies of themselves, wouldn't they be alive?

July 29, 2014 | Unregistered CommenterMax

I believe that machines can never be conscious. I believe they will soon be able to pass this test and I think that's great. But a machine will never feel as a biological entity feels, even if it can think in a parody of how a biological entity thinks.

July 29, 2014 | Unregistered CommenterChristopher Penny

How appropriate that posting comments here requires solving a CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart).
Instead of the old distorted text, I'm now getting photos of house numbers from Google Street View.
In April, Google said that their Deep Convolutional Neural Networks had 90% accuracy in detecting and reading difficult numbers in Street View, and the same algorithm could solve the hardest distorted text puzzles from reCAPTCHA with over 99% accuracy, better than some humans I know.

Other websites have CAPTCHAs that require answering simple questions, which ought to be easy for a chatbot that can pass a Turing test.

July 29, 2014 | Unregistered CommenterMax

Greg, that's not a good comparison.

A paper airplane and a 747 are different things - they may share the concept "airplane", and we can say correctly that "those are both planes", but while that's acceptable for language it's not acceptable logically. We understand when that is said that the paper airplane shares only the concept of airplane with the 747.

For example, you cannot get into a paper airplane. A paper airplane doesn't perform the functions of a 747, and isn't even a good model for it's larger counterpart.

So it is strange to make the claim that because you can create a facsimile plane out of paper, the material a brain needs to have a mind should be various. You will have then created something that shares characteristics with a mind, but falls short given that it is not the same type of thing, but merely shares conceptual features.

Further, natural selection processes are not optimal and, importantly, are not really predictable. Organisms are not finely tuned, and the features an organism carries forward don't only depend on how successful they are as tools for propagation.

Look at something like language, which may tie into intelligence, or the type of consciousness we associate with humans. It may have evolved entirely as an accident, as there is evidence to support the idea that humans had vocal capability and sufficient communication tools available before language was fully adopted. Language may have been a byproduct of a need for an increased brain size, and of course the organism which developed it would grant intelligence primacy as a survival mechanism which may not be the case.

Given that intelligence arose, seemingly, in only one species and very recently in that creature's development, it's very sensible to have a bias for thinking that whatever we call human intelligence is based in biology. Well, it's a matter of fact that biological processes developed it, but it's also reasonable to think that even a material such as silicon or man-made brains-in-vats might not sufficiently mimic the product of the evolutionary process.

July 29, 2014 | Unregistered CommenterNick

Max:

"What's so special about the brain that can't be reproduced with an artificial neural network?"

Because we're not sure that a mind is fostered by neural networks, or for that matter that this is even how the brain actually operates, even though we may see connections which we would rationalize as a theory for neural networking. As Massimo said, we're still not entirely sure what it is we're looking for - what an intelligence/mind/consciousness is.

So it's not just a "what's so special?" It's a "We can't even definitively confirm that we understand what we are trying to reproduce"; this is why I am personally so skeptical of modern claims for AI. And it certainly was not what Turing intended to say in his 8 page paper which spawned all these contests.

July 29, 2014 | Unregistered CommenterNick

Massimo makes a great point about judging whether something is conscious by comparing it to other things that we consider conscious.

However, consider the thought experiment where a person's brain function is fully simulated in software. Above argument is thus forced to claim that consciousness goes beyond brain function, or say that brain function cannot be simulated in software.

July 29, 2014 | Unregistered CommenterAlexey

Nick,

You wrote: "Because we're not sure that a mind is fostered by neural networks, or for that matter that this is even how the brain actually operates.."

At this point we know with very high certainty that things you listed - intelligence, mind, consciousness are produced by the brain. These are not magical mysteries but empirical questions. Let us not lose sight of that fact.

July 29, 2014 | Unregistered CommenterAlexey

Nick,

There's a lot of biology on the planet, yet only brains are conscious. So there has to be something special about the biology of brains in particular. Well, we know that brains have all these connections that are like nothing else in nature, but this can be reproduced artificially on an analog computer.

Alexey,

I've long argued that a brain simulation on a standard computer would behave just like a brain, but it would not be conscious, because if you look at the actual hardware, it's just doing one simple arithmetic calculation per clock cycle and stores the result in memory, where it sits like text in a book. Text in a book isn't conscious. A calculator doing arithmetic isn't conscious. I just don't see at what point the system is conscious. I think it would be the so-called "philosophical zombie," which is fine with me, because I want to be able to turn it off without committing murder.
In contrast, if you look at what's going on in a brain, I can imagine that certain patterns of activity ARE sensations and feelings.

July 29, 2014 | Unregistered CommenterMax

Max,

You wrote: "I've long argued that a brain simulation on a standard computer would behave just like a brain, but it would not be conscious, because if you look at the actual hardware, it's just doing one simple arithmetic calculation per clock cycle and stores the result in memory, where it sits like text in a book. ...A calculator... I want to be able to turn it off without committing murder"

This works if you define consciousness as something that can only be done by a brain.

Consider a thought experiment where you are not dealing with a calculator but a real body. It is running software, but it was created to fully replicate the mental life of a recently deceased human being. It looks and acts like a real human and it is begging to be kept alive. Can you still turn it off without committing murder?

July 29, 2014 | Unregistered CommenterAlexey

Alexey,

Well, I wouldn't feel sorry for the android because, as I explained above, it doesn't actually feel anything or experience fear of death, even if it acts like it does. If your computer, instead of saying "Are you sure you want to shut down?" said "No, please don't shut me down!" would that make any difference?
But I suppose if we get to a point where there are a lot of these androids running around behaving like humans, turning them off willy-nilly would probably have the same detrimental effects on society that murder does, so it would make sense to outlaw this, but maybe call it something other than murder.

July 29, 2014 | Unregistered CommenterMax

Max,

How do you determine whether an entity actually feels something?

July 30, 2014 | Unregistered CommenterAlexey

Alexey,

I don't know how to determine it in general for any entity, but I already explained why I think a Turing machine doesn't feel anything, because it's essentially a calculator with memory, or technically, a tape and an instruction table with a read/write head. I doubt that some special sequence of arithmetic operations not only simulates feelings but IS feeling something.

July 30, 2014 | Unregistered CommenterMax

"For example, you cannot get into a paper airplane. "

Nick:

You're picking arbitrary nits. Two things that are alike are always different in some respects. The characteristic that Julie identified as characterizing an airplane is "flight", which a paper airplane can do just fine.

July 30, 2014 | Unregistered CommenterGreg Esres

"yet only brains are conscious...A calculator doing arithmetic isn't conscious. "

Max: You can't know this. Without knowing what consciousness is, any statement about what does or does not possess it is pointless.

July 30, 2014 | Unregistered CommenterGreg Esres

Greg, you think books and calculators might be conscious? Do you worry about shutting down your computer because it might be conscious?

July 30, 2014 | Unregistered CommenterMax

Max,

There is not much to discuss if you cannot state the general rule you are using to determine whether something is conscious.

Your point that a simple machine is not conscious does little to advance the discussion about complex machines. I can easily grant that simple machines are not conscious just like i can grant that simple animals are not conscious.

July 30, 2014 | Unregistered CommenterAlexey

On the question of self awareness, in biological entities this has nothing to do with intelligence. In the podcast you imply that with increasing intelligence self awareness will arise. But if we think of an insect, with arguably a very low intelligence, it is still self aware. If you suddenly put something in front of an insect it will turn and head the other way. What is it thinking? I propose that it thinks 'I am at risk- I will take action to reduce the risk to myself.' It is aware of itself and can risk assess. Machines, no matter how intelligent will never be able to do this because they need to be able to impute a self onto their body and mind, and they cannot make this mistake, unlike biological entities.

July 30, 2014 | Unregistered CommenterChristopher Penny

Alexey,

All right, you grant that simple machines are not conscious, and my point is that a Turing machine is one such simple machine whose architecture is too simple to be conscious, but it can execute a complex brain simulation that behaves like a brain. That's a big deal. As I recall, Massimo was skeptical that such a thing is possible.

July 30, 2014 | Unregistered CommenterMax

Max,

I understand that Turing machines can have arbitrarily complex architectures. Consider a thought experiment where a Turing machine fully replaces the function of the brain. Some people have real brains and some people have Turing machines. How would you decide which person has consciousness?


Christopher,

You make a good point about differences between intelligence and self-awareness. However, I would caution against using insects' capabilities to judge their thinking and self-awareness.

July 30, 2014 | Unregistered CommenterAlexey

Alexey,

I'm thinking of the simplest Turing machine that can simulate a brain. Because it behaves like a brain, it shows that behavior alone doesn't tell you whether something is conscious. It also raises the question whether consciousness is just an evolutionary fluke, since intelligent behavior can happen without it.

July 31, 2014 | Unregistered CommenterMax

Max,

Consciousness is produced by the brain... therefore one may argue that a Turing machine which simulates the brain would necessarily produce consciousness. What do you think about that?

July 31, 2014 | Unregistered CommenterAlexey

PostPost a New Comment

Enter your information below to add a new comment.

My response is on my own website »
Author Email (optional):
Author URL (optional):
Post:
 
All HTML will be escaped. Hyperlinks will be created for URLs automatically.