How Likely Is It That A Subset Of Humans That Can Pass A Turing Test Aren't Conscious?

How likely is it that a subset of the 6 billion people that can pass a turing test aren't conscious? That they don't have subjective conscious, but are fooling those of us that do? I suspect the odds reasonably would be greater than zero -and perhaps significantly greater. This is the flip side of the argument "if a computer program or an upload can pass a turing test, then it's subjectively conscious". Perhaps some of the 'people' we thnk are subjectively conscious because they pass our current, informal turing tests are not.

Thoughts?

Note: 4/27/08 I just corrected the wording in the title and post to correct a typo that's been in this post for about 6 months. Of course I meant "that can pass a turing test aren't conscious", not "are turing machines". This was caught early in the comments, and I think the rest of the post made it clear what I meant, so I was in no rush to correct it.

Comments

[Sebastian Hagen]

What do you mean by the phrase "subjective conscious"?
How do you tell agents that have it apart from those who don't?
Assuming you are talking about a concept that describes an aspect of reality, you should be able to make testable predictions conditional on its presence or absence in an agent.

Once you do, testing your hypothesis that this quality is not universal among humans becomes a relatively straightforward matter of picking a number of humans and testing for the predicted symptoms.

Personally, I just don't understand what you are talking about. Since you mention turing machines, does your concept have something to do with Hypercomputation? No humans have ever been shown to be able to perform any non-turing-computable operations.

[TGGP]

I'm not really a wonk on the subject, but "consciousness" doesn't seem to be that meaningful a concept to me, other than in distinguishing the awake from the sleeping or knocked out. What expected experiences would you have with a human turing machine compared to a conscious one?

Just noticed that Sebastian Hagan is saying the same things. Great minds think alike and whatnot.

[Nick Tarleton]

I believe the standard phrase for these people would be philosophical zombies.

Why do you think this is at all likely?

[Hopefully Anonymous]

TGGP, my experiences with them may not be different, but if it's true I think it would reduce the moral claim (among us non-zombies) to save all people currently living from death. A consensus that fewer people need to be saved may help make it more likely that the people we attempt to save actually do get saved.

Nick, I'm not sure how likely it is, but I think it should be on the speculative table, since others are suggesting anything that passes a turing test isn't a zombie.

[Hopefully Anonymous]

p.s. Nick, start blogging. Sebastien you too, please.

[TGGP]

You didn't answer the question about how to distinguish humans from zombies. Go ahead and assign a moral weight of zero to everyone other than you, but don't try to bring in unfalsifiable stuff about machines and zombies.

[Hopefully Anonymous]

Actually, I think the approach that assumes that no humans (above the age of four, and appearing to be in the normal range of cognitive health) to be philosophical zombies is bringing in stuff that's unfalsifiable with current scientific ability.

I think our default approach should be openness to the possibility that a subset (from 0-100%) are philosophical zombies. That's a good starting point to bring in the empirical inquiry, in my opinion.

[Nick Tarleton]

Actually, I think the approach that assumes that no humans (above the age of four, and appearing to be in the normal range of cognitive health) to be philosophical zombies is bringing in stuff that's unfalsifiable with current scientific ability.

So does any position on the zombitude of others.

Occam's Razor, and the fact that all humans are pretty similar behaviorally and structurally, and the fact that I have experiences, lead me to believe that all apparently conscious humans have experiences. Confirming or disconfirming evidence would be nice (and I expect cognitive scientists will eventually come up with some) but I feel fairly confident where I am.

[b]

if you're really interested in the literature of zombiehood, or what have you, read dennett's 'sweet dreams: philosophical obstacles to a science of consciousness'. my brain is too porous to remember his arguments, but he devotes some chapters to demolishing zombiehood, along the lines of sebastien's comments.

[Hopefully Anonymous]

b,
Thanks for the reading recommendation. I'm skeptical anyone knows enough to create arguments that successfully "demolish zombiehood".

Nick, I'm not sure Occam's Razor (which I don't think is a natural "law" derived from empiricism) would lead us to think that no humans are philosophical zombies. Subjective consciousness (and the intersubjective "space") still seems to me to be too weird -or magical, as Eliezer Yudokowsky would say- to be situated within the Occam's Razor framework, it seems to me.

[Nick Tarleton]

I am inclined to take Eliezer's advice against postulating magic in consciousness. Certainly the fact that I have phenomenal experiences is remarkable and it's hard to imagine how it fits into a physicalist framework, but as Scott Aaronson likes to say, even though it's hard to imagine how a physical object could have experiences, it's no easier to imagine how a nonphysical, um, thingy could. Postulating a whole other domain of reality just so I can shove one mystery (I don't see anything additionally mysterious about intersubjectivity) into it - making no retrodictions, very few predictions, and not clearing the mystery up any - is not parsimonious.

I also think (although I don't know how rigorous this argument is) that even if there were a nonphysical part of the mind, the idea that we could discover it through introspection makes about as much sense as the idea that an AI could discover through introspection "my processors are made of silicon", in other words none. How's the brain to know the difference between physics and magic that easily?

Occam's Razor is not an empirical law (neither is Bayes' Theorem); it's the informal version of Solomonoff induction.

One simple argument that I think is a strong blow to zombiehood: By assumption, the zombie is behaviorally identical to a human. Human behavior includes discussing consciousness. Therefore, a zombie can discuss consciousness exactly like a conscious human. Therefore you would discuss consciousness whether or not you were a zombie. But that would mean that your discussion of consciousness cannot be caused by your being conscious. Reductio ad absurdum, QED.

[Hopefully Anonymous]

Nick, your paragraphs #1 and 2 seem to be based on a misunderstanding of the last sentence of my previous post.

I'm using "magic" as an intentional placeholder, as Eliezer recommends in this post:

http://www.overcomingbias.com/2007/08/say-not-complex.html

"Marcello and I developed a convention in our AI work: when we ran into something we didn't understand, which was often, we would say "magic" - as in, "X magically does Y" - to remind ourselves that here was an unsolved problem, a gap in our understanding. It is far better to say "magic", than "complexity" or "emergence"; the latter words create an illusion of understanding. Wiser to say "magic", and leave yourself a placeholder, a reminder of work you will have to do later."

[Hopefully Anonymous]

Nick, your last paragraph doesn't seem to me to be a strong blow zombiehood at all. Instead it seems to point out the theory, which I find to be most salient, that our ability to determine what's a zombie and what's conscious is limited by our bounded rationality and limited empirical abilities.

[TGGP]

What does zombiehood bring to the table? What experiences does it cause us to anticipate? It seems to me the answer is nothing and I do not see how the concept is of any use. Don't go on about how it is "unscientific" to "assume" healthy people above the age of four are not zombies. I could make up a term, like "fnarl" to denote people who are "fnarlish" as opposed to "unfnarlished", and then wail like the TimeCube guy that we are educated stupid to assume nobody is fnarlish, when it is simply the case that it is a useless concept we do not need to contemplate and makes no difference if it is true or false.

[Hopefully Anonymous]

Well, as a policy matter, zombiehood oould bring something significant to the table: it may facilitate more selective policy about who to maximize persistence odds for. We'll likely do better maximizing the persistence odds for a subset of humanity than all of it. As a policy matter, we might do better with a rebuttable standard that humans aren't subjectively conscious.

I understand that that's different than what the best science tells us. I agree that you may be right that our study of consciousness is primitive to the point that one can't posit zombiehood (or non-zombiehood). But non-zombiehood doesn't seem to me to be the default position. The default position is a question mark, as best I can tell. It seems to me to be fake knowledge/fake explanation to go beyond that an posit an intersubjective space (or the lack thereof) for all humans, or all phenomena that act like or momentarily convince us that they have an intersubjective space.

But it may be wasted energy for us to think much about that now, other than as arguments against "leave no human behind" policies.

[TGGP]

"Well, as a policy matter, fnarlishness could bring something significant to the table: it may facilitate more selective policy about who to maximize persistence odds for. We'll likely do better maximizing the persistence odds for a subset of humanity than all of it."
So how do we decide if zombiehood or fnarlishness is more important?

[Hopefully Anonymous]

TGGP, as a policy matter? I think you can predict my response:

Hint: it's the second clause in my blog description.

[TGGP]

Does understanding zombiehood or fnarlishness help you to live forever in any way other than convincing people to prioritize you over the zombies and fnarls?

[Hopefully Anonymous]

TGGP, yes. A zombie version of me would be fine for you but failure of mission for me. Weak versions of zombies of me can already be constructed. How weak or strong a zombie version is seems to me to be a technological race between detecftion and deception. I understand zombies: observers may think it's me but I don't experience subjective consciousness through it. As for fnarlishness, either it's a synonym for zombie, or it's not. In the latter case I think you're wasting energy keeping "fnarlishness" in play in this discussion.

[steven]

It seems that by "Turing machine" you mean a non-conscious machine that passes the Turing test. That's not what "Turing machine" means: http://en.wikipedia.org/wiki/Turing_machine

[Hopefully Anonymous]

Steven, thanks, I was aware of that (and I'm sure other readers are too) from the first comment in this thread. I'm keeping the title just for honesty/transparency.

[Hopefully Anonymous]

This blog (and my activity on other blogs) will be greatly reduced for the next 3 months or so, to get quite a bit of work done in my real life job. But then I plan to return to current levels of activity.

[TGGP]

"Fnarlishness" is a concept I haven't defined yet, but only introduced. I bring it up as a hurdle for the zombie-concept to overcome. If the zombie concept does not help us to understand reality any better than fnarlishness, we should be indifferent between which we want to focus on. In your comment you do seem to bring something new by using zombies as a simulation for or duplicate of something else that already exists. So is your question whether there are some groups of people that we believe to be one but actually have more than one body?

[Noter]

HA,

You want to engage in 'empirical inquiry' to figure out who the zombies are? The definition of zombiedom precludes the success of such inquiry.

[Noter]

Also, what policy implications exactly? You don't care about the welfare of others irrespective of whether or not they're conscious. If for some reason you decide that left-handed people have no subjective consciousness, they will still act as if they do and respond poorly to advocacy of policies that ignore their interests.

People do very in their attention allocation and level of automatization (not remembering driving to work, etc) and you might be able to measure variance in that, but it wouldn't be different in kind from saying that we should treat the lives of happy people as more valuable than those of the depressed.