Yes, all the 50+ year Turing test literature is surely relevant AND everything anybody has ever written about consciousness! Maybe have a look at some of Daniel Dennett's writings for a start if you haven't already -- not that I think he really gets it right...

@Andrew: I agree that defining sentience would be a good first step, if possible; but it may be as impossible as passing the Turing test to everyone's satisfaction.  (In the decades since I first heard of the latter, the terms of reference have shifted every time a computer has passed the current version of the test.) 

Nevertheless there seem to be quite a few undefinable qualities for which widely accepted criteria have been described.  Love, for instance.  Perhaps we can still hope for some battery of "sentience" tests that would satisfy almost everyone.  Or maybe acceptance will just take place gradually.... 

I disagree with your implication that understanding phenomena tarnishes their lustre.  I know that blue skies and red sunsets are caused by Rayleigh scattering, but for me that dims their magic not one iota. 

Jess,

Fascinating question.  I think that as time goes by our definition will change as it always has.  Our understanding only makes us wonder if we are correct in the current definition.  As we become more educated and learn how wrong we were in the past about education and science we tend to grow our definitions around the new understandings and this means we change.  Is this a case of us just catching up with reality or are we really different than we thought?

Science is not sentience so is our human form the only thing that is?  Sounds like a Philosophical question.

Certainly it is a Philosophical question.  (Are any other kinds worth asking?)  But is it a stupid, pointless excuse for bickering over undecidable opinions, like so many Philosophical questions?  If I thought so, I wouldn't have asked. 

In the spirit of full disclosure, I will reveal my own opinions:

Computers (I use this term in its broadest sense, not just "programmable" computers) will gradually master every version of the Turing test we reinvent, until everyone is forced to admit that computers are able to simulate intelligence, self-awareness, acculturation, passion, empathy and any other ineluctable quality we claim as part of sentience.  Then the question will come down to, "Do simulations count as real?" which is a truly important Philosophical question. 

I contend that huge swathes of human "sentient" behaviour are in fact simulations that we have learned in order to function in a society with huge swathes of misinformation masquerading as "given truth".  In fact one could argue that sentience itself is the ability to exhibit such simulations on demand.  (But I'm jumping ahead.  Forget I said that....  ;-)

Jess,

You are correct it is a very important question and needs to be asked.  There is a larger issue here.   Could these simulations that we are creating be the only way that we may survive outside the life of our star the sun.  If we shoot these programs into the universe to explore for us and then our planet gets consumed by the sun in a few billion years or sooner if there are other events that happen.  Will this be the only way that some other advanced civilization would find that we ever existed and be able to recreate what we were like?

It sounds like science fiction but with the understanding of space travel that we have today it is our only hope of reaching other worlds.  If that is the case we need to make these simulations with great care and use them wisely.  If we do not others may think of us as very shallow or not very intelligent at all.

The idea that sentience is only being able to produce the correct response is a programing issue in computers just like it is a programing issue in people.  The real problem is that computers unlike people will really mean what they say and there fore could be considered more life like than people that say one thing but behave in the opposite way when no one is looking.  I guess that can be programed in as well.

Almost anything can be programmed in a computer if you have enough memory.  One of my favorite comity lines is "you can't fix stupid" and I think this is the problem we will have with sentience in machines is that we can not stop the stupid programing that will happen with any technology.

I am certain that we will have machines in the future that think they are alive and therefore we will have to deal with that as it happens but I am not so sure that we as a society will do a good job of being less stupid about how we treat each other let alone sentience without flesh and bone.   

I'm certain we will soon have computers that will say they think they are alive, same as we now have people who say they are prioritizing the public's best interests.  If we can devise a scheme for determining the veracity of the former claim, perhaps we can apply the same criteria to the latter. 

As for far-future aliens, I don't really care what they think of us unless "we" are still around. 

My whole thought is that it would be nice to think that our existence is truly with meaning and will survive.  That being the case I want to have the best shoot at making it look good.

Jess,

I do not have any illusions that I will be alive but maybe some distant relative and that is comforting.

@Francesca:   Uh... you sure have a lot to say, filled with passion and narrative.  Have you considered writing a Book? 

Hi Jess - - - for starters, you might want to check the Turing Test and work forward from that using the citations index.  He asked the same or similar question and proposed a test.  Take a look at it: http://en.wikipedia.org/wiki/Turing_test

Hope this helps,

Jerald

@Jerald:  As mentioned earlier, in the years since Turing proposed his test for AI in 1950 there has been a stepwise retreat from his initial version as computers "passed" each iteration of the test.  Each time, proponents of the test said (in effect), "No, no, that's not what we meant.  That doesn't count!"  I have been following these dodges since the 1970s and by now I'm pretty sure there is no way the Academy is ever going to declare a computer to have achieved sentience, intelligence or any other cherished human attribute, based on any version of the test.  It's fun to watch the spectacle, though. 

I should really have asked whether Simulations are Real; that's the deeper question.  Greg Egan has written some excellent novels exploring just this issue. 

Might want to think about and sharpen the definition of sentience - - - all of the tests assume communication as an essential part of the definition as that's what seems to be center in the tests.  AI got into this sort of trouble as everyone seemed to have a different idea or definition of intelligence.  Once you have a tight definition with testable components then you have a chance, otherwise, it's just interesting conversation.

Just interesting conversation?  Methinks your expectations are extravagant, Jerald.  But I'll play along.  The typical approach to sharpening the definition of sentience would be to make a list of other undefinable qualities, all of which must be exhibited by a sentient entity.  A short list might be "general intelligence", "self-awareness", "apprehension of [one's own] mortality" and "ability to communicate".  Each of these is deserving of its own interminable discussion; but do we really want to go there?  Can't we just treat "sentience" like "love" or "evil" and say, "You know what I mean"?  Even without a sharp definition we can ask ourselves what it would take to convince us that our "smart phone" was actually sentient. 

I still think the deeper philosophical question is whether a perfect Simulation is distinguishable in any meaningful way from the Real Thing.  If my smart phone acts sentient, will I treat it as sentient? 

Good thinking -  - - and if the simulation is "perfect" - then by that definition it would pass.  That might leave out all "things" that have a self awareness and do not communicate.  It's always the words that get us into trouble.  For example, suppose that dualism is the model and the mind/brain is just the receiver.  Then, with a perfect simulation, we might be in trouble.

Yeah!  Many people say, "It doesn't matter" (whether my own sentience [or whatever] is "just" a simulation), but within the context of Animism and/or Dualism it matters a lot!

Agreed!

@Rene: I try not to AssUme anything, but of course I fail.  :-)   The question is, can we assume (or deduce) any reason why conventional computers cannot become sentient?  And how about unconventional artifices like neural networks? 

An ongoing debate with my friend Georgia pits (to put it glibly) my Instrumentalism against her Animism: I am most comfortable when I can come up with a satisfactory materialist mechanism for each factor in sentience, thus denying the existence of Chalmers' "hard problem" of qualia.  Perhaps we just imagine that sensory perception changes to something different when it "enters our consciousness".  She counters that perhaps we just imagine that intelligence/consciousness/sentience is a product of material instruments; what if we are merely antennae tuned to the soul of the universe, so to speak?  (Francesca will like this, I bet!)  I find myself questioning the wisdom of my commitment to Instrumentalism (which sounds like a derogatory term, but I have to admit that's what it is). 

We may live in the epoch when these questions will be answered, or at least clarified.  That's pretty exciting, and not a little scary.  Can a sufficiently powerful computer become sentient?  Have a "soul"?  We'll have to ask them!  (And, for that reason, be careful not to write it into their programming that their main purpose is to fool us!)

Francesca, I find your posts interesting and thought-provoking, but there are a couple of things you do that make them difficult for most Instrumentalist and Academic types to read with an open mind.  The capitalization of nouns is only a little disconcerting, and can be easily accommodated by those who have studied German; but your propensity to make categorical declarations is a big problem.  We are used to hearing statements like, "I believe we are all connected," but statements like, "We are all connected," belong in church or other places where contrary arguments are not appreciated. 

Yes, I know, I do it too, but I try to maintain the humble academic charade for as long as I can; I think the conventional quota is something like one pronunciamento for every five or ten conditional suggestions with caveats.  Some people seem to be able to remain open-minded indefinitely, but they can get annoying too; yesterday I coined the phrase,"ostentatious humility and patience" to describe this affect. 

Anyway, see if you can speak with slightly less absolute authority; it will make everyone more comfortable, even if it's only a Simulation.   (See, this is on topic, sort of.  :-)

Hi Werner!  Certainly supervenience (I had to look it up, but the idea was immediately familiar) has always been the physicist's guide, although there is considerable debate over whether complex phenomena "reduce" to simple phenomena -- see for instance Phil Anderson's nice essay, "More is Different" in which he makes a great case for "emergence" of genuinely new phenomena from "mere" complexity.  So I tend to think just the way you do; but I'm trying to learn to question the convictions of a lifetime of Instrumentalism.  (It's no picnic -- I wonder, "Will I still be me if I relax my grip on my most important paradigm?") 

Ramachandran's essay is similarly grounded in supervenience -- this is just the way we academic scientists have to think. or none of our colleagues will listen to anything we say. 

Worst of all, he/you/I/we may turn out to be right!  Maybe all we need is supervenience (augmented by emergence) to understand everything we can understand, or need to.  But what if... oh, never mind, I can't even say it without feeling silly, and a little scared. 

I'll probably regret posting this, but what the hell.

Words like'sentience' worry me. Somebody once said "When the conceptual analysis is over then the philosophy can start.". I think this is right but it glosses over the possible need for conceptual analysis before starting philosophy.

 Humans began to talk because they had something to talk about and someone to talk to. This is back to the frame problem of AI and this is hard only because we confuse being 'intelligent (sentient)' with being like us. Whatever tests we set up for sentience we will be left with this dilemma of whether sentience is some measurable capacity, which will then be a challenge to replicate or some observable distinction between species which we can name or just being like us.

I have found it helpful to define two sorts of concept; linguistic concepts for those entities that we talk about and animal concepts for those entities that have a place in our world but we do not share. Defined like this, I suggest that some other species seem to have animal concepts since they seem by their actions to give a place to things in their world. Neurally,this requires integration of perception into holisms of recognition and the capacity for combinatorial processing of these. See: Lawrence Barsalou, (1999), Perceptual symbol systems, B&BS, 22, 557-660. Same idea, different name.

I suggest that this could be termed sentience. Dogs and chimps and some birds seem to qualify. Devising a test should be a job for the ethologists but there is a risk of checking for whether an animal has a human concept rather than having its own.

Hope this helps

@William:  All good points. I particularly liked the idea that sentience (which I cannot define satisfactorily, but some others might attach to "self-awareness" as much as "intelligence") arises from having something to say and someone to say it to -- i.e. communication among peers.  This has implications for the Internet, I suspect....

I also like your "integration of perception into holisms of recognition" which I tend to think of as "generalizations" -- projections of complex and usually unique qualia onto abstract shadows that we give names for the sake of communication.  Sort of the reverse of Plato's ideal chair, which contains all possible chairs: the word "chair" is a mere shadow (projection) of our actual perceptions of chairs onto an abstraction simple enough to define.  And then of course there's the conundrum of how well the actual chair (assuming it exists) corresponds to our perceptual qualia that we project onto "chair".  

Here's some thoughts I wrote out on this subject some years ago:  http://jick.net/~jess/hr/skept/YinYang/

I have to confess I do not know what "B&BS" stands for, although some possibilities come to mind.  Can you expand the acronym for me? 

Dear Jess,

"until everyone is forced to admit that computers are able to simulate intelligence, self-awareness, acculturation, passion, empathy and any other ineluctable quality we claim as part of sentience."

Sorry but you are wrong with this. They won't ever be able to do it because we will never be able to define such abilities. the book  the "Descartes' error " tells you scientifically that it is not even thinkable to approach feelings to a machine because our feelings are not from our brain..

http://en.wikipedia.org/wiki/Descartes'_Error

If you want to understand sentience it is not a case of 'either or'.  Sentience is graduated and subject to limitations.  If it were just self awareness or awareness of surroundings then veritably all animals and plants would be sentient since they all react to stimuli.

We acquire sentience and develop it.  Chimps do to a lesser degree, dolphins lesser still but we and them all have it.  Even bacteria have a form of sentence. They 'know' what to avoid.  Eukaryotic cells in multi cellular organisms have sentience.  Just watch how 'drug aware' tumor cells efflux chemotherapy drugs and signal to other cells that there is danger. 

@Barry: Glad you agree with the premises in http://jick.net/wordpress/?p=166 !  ;-)

There is a practical test that I call the inverse-Turing test. A sophisticated artificial person may throw off an array of clues in their detailed carriage, just as we do. If they do, a typical real human given plenty of time with the thing would most likely judge whether the thing was actually feeling (or had at least been constructed as such a nearly identical copy of us that it appeared so). It would be hard enough to make such a mind in the first place, so a deliberate effort by either it or its maker to pose in such detail so as to fool would be an even bigger job to get right and probably would fail for a long time. So even though the this test could fail in theory, it would most likely work, using our same circuits that can tell whether another person is sincerely feeling something.

@Karl: See https://www.youtube.com/watch?v=PdJm3DVg3EM for a poignant counterargument. You are probably right for now, but these sims are progressing really fast; and combined with Machine Learning (from past failures) I wouldn't be surprised if the Turing test became useless within 5-10 years. Time together may help, but have you never met a human whose consciousness seemed consistently absent? ;-)

@Jess I think we are in large agreement, but for 1. You might expect faster "progress" than I, and 2. Some folks are purposely trying to make their machines very convincing just as an end in itself, just as there are some people who are excellent actors and choose to act, whereas most folks are worse or-less inclined liars. It is certainly an imperfect test, but most real world useful ones are. Conveniently for me, most of the humans with consistently absent consciousness don't hang 'round the consciousness-ponderers water cooler.