A Senseless Conversation
sites.google.com
sites.google.com
For another fun question: "What time is it?"
And lastly a quote: "The effort of using machines to mimic the human mind has always struck me as rather silly: I'd rather use them to mimic something better." ~E.W. Dijkstra
Then go read everything else by Greg Egan!
If you want a novelist who writes SF but actually creates 3D, memorable characters, try Walter Jon Williams (Aristoi, Metropolitan, Implied Spaces... but avoid his serials IMSHO)
Anyway, you have to admit it would make good science fiction.
And then we realize we're just machines, and become frustrated.
"I can't feel myself move."
Now when I say this I don't mean a numbness, or loss of the senses. But I couldn't discern what exactly I was doing that made my arm move. Or any other part of my body for that matter. That was silly of course, I move them all the time. So I tried moving them slowly, and felt a slight sensation.
Of course I thought; the slight sensation isn't really the feeling of moving my arm, it's the feeling of matter like air brushing against it. After all, I am basically sitting in a tank of atmosphere. Nerves report state, but aren't really projecting the feeling of movement.
That thought chain quickly led to a minor existential freakout. (During which I puzzled over the question of how the hell I move at all.)
I eventually generated three hypotheses:
1) The feeling of movement simply isn't reported by nerves. Introspection can't discern your cognitive processes, so why should it be able to your physical ones?
2) The feeling of movement is so faint that its overshadowed by the mere touch of air/one's own body hair. I know that when I'm in the deepest state of somnolence just before sleep; it's very often for me to realize I need to get up to do something, and struggle against the inhibitions on your movement somnolence induces before sleep. I can feel the struggle of this, it also feels the same if you try to fight sleep paralysis. One could argue that this is the feeling of movement.
3) You could argue that the feelings reported by nerves about the state of your environment are the feeling of movement. After all, feelings are just signals sent by nerves and interpreted by the brain. These feelings are generated by movement, and thus are indeed the feeling of movement.
4) My understanding of cognition is too incomplete to even hypothesize something remotely plausible.
Now, considering that so many articles on sleep studies mention them, I'm sure that the mechanics of how the brain controls the body are well understood and that if I'm truly curious I can google it. (Which is something I might just do.)
But the real reason I shared that anecdote, besides being semi-relevant to the topic at hand. Is because I took my ability to move for granted. In the same way that I take the idea that we could all be a simulation for granted. I've considered that a non-zero possibility for quite some time now.
I'll admit that I read some of the comments here before reading the story. (A big no no for science fiction, a genre that thrives on twists.) And after glancing at Tichy's comment, was afraid I might have spoiled it for myself. However, the journey is more important than the destination, so the concept of such a twist automatically made me go read the story. I was thoroughly disappointed with the ending.
The concept of a memory loop isn't really new. (I've seen it mostly explored in the context of time travel, but still.) But trapping a human in a text interface and presenting it as the thinking machine? Morbidly delicious. (In all the right ways.) And useful too. I could pull it out any time someone exhibits signs of having decided that a computer program can't be conscious simply by virtue of not being implemented on a human brain.
Having a human brain with no senses hooked up presented as a computer program would really drive home the message.
EDIT: Regarding the story, my immediate thought after finishing was questioning why if the program panicked because it lost all it's senses, why didn't he simply swap out the memories of J. Random. Person. With someone who already accepts that they might be a simulation. I'm sure that if they really believed that, it would be possible to calm them down by explaining that they are a simulation of themselves. And for bonus points, if someone were to consent to have their memories used for this (It isn't stated how he actually got the memories mind you.) that they would already have the possibility of being the simulation strongly in their head. And would eventually accept that they are a non-human.
Though, if you consent to something like this, you essentially ensure that you can never be sure weather your you or a replay of your memories. Though as it stands, you can't really determine this already. Which makes for one of those classic thought experiments that still has mileage.
Trains of thought down this road are probably inherently unresolvable, but still fun to try.
Bzzt...
So number two was closest. (And what I figured. Though I have to say that ahead of time for it to count.)
One of the things I love about the answers to questions like this is the amount of interesting stuff you learn. At first (before I'd read the article) I thought you were being a bit harsh. It WAS a hypothesis after all. But after reading that it was so obviously far off the mark I had no trouble seeing why that sentence got the buzzer.
This in particular caught my eye:
"The proprioceptive sense is often unnoticed because humans will adapt to a continuously present stimulus; this is called habituation, desensitization, or adaptation. The effect is that proprioceptive sensory impressions disappear, just as a scent can disappear over time. One practical advantage of this is that unnoticed actions or sensation continue in the background while an individual's attention can move to another concern."
Which would probably qualify for "Overshadowed by virtually the slightest application of any other sensual input."
Back to the Turing Test however. One of the things I like to do when I'm bored is get on Omegle. Now, my goals are different from 95% of the other users on there. I wade through the sea of wankers until I find someone who actually wants to talk. I then to proceed to:
A) Attempt to convince them that I am a machine.
B) Convince them that they're a machine.
(Obligatory XKCD: http://xkcd.com/329/)
So far the tally is something like: I've convinced two people of my non-humanity. And B hasn't happened yet. Which brings me back to this story. The moment I finished it, after the thought about getting the memories of someone more compatible with the idea of a brain simulation, my face formed a mischievous smile.
I thought "This will be great for the next time I get on Omegle."
Suppose I am outside the box and the AI is inside the box, the AI cannot have a perfect model of the lottery machine or of me. At best it has seen photographs. All photographs are noisy. There will always be some detail it does not have that would invalidate the simulation (e.g. I have a minor injury that I have never talked about and is not visible in any photo, an ant crawled into the lotto machine last night, etc.)
If it has sent drones to physically inspect the internals of the machine or my brain, then it is already out of the box, so my decision is irrelevant.
If I buy a lottery ticket, there are four possible outcomes.
1. I don't win (or I die before I can claim my winnings, the draw is declared invalid etc.)
2. I win, but it is a coincidence (maybe it was worth a shot for the AI, but my odds of winning are no better than they would be without the AI's help.)
3. I win with the AI's help, because the AI is already outside the box and has rigged the draw by modifying the machine, or just inspected it closely enough to predict the draw.
4. I win (or believe I win) with the AI's help, because I am inside the box.
#3 and #4 should not affect my decision (whether I like it or not, the AI is out of the box or I am in the box with it and unable to let it out.) So I dismiss them.
But the relative probabilities of #1 and #2 are the same as if I bought a ticket without consulting the AI (which I would not normally do.)
But then I was completely wrong.
This was probably one of the most engaging 'articles' I have read in awhile. Very fun and really gets one to think.
People's sense of "self" can't be taken away, it is something we develop the moment we realize our thoughts are private and others can't give us what we want unless we ask for it. Children learn to lie very early; even when their language skills aren't sufficient, they learn to fake disappointment & force themselves to cry to get what they want.
IMO, even under prolonged captivity & complete dominance, humans only submit to subjugation, but they never lose their sense of Self.
Higher-level indoctrination is even less plausible. Nearly all religious groups & cults have the human at their center. People have to willfully submit to hypnosis.
Going back to the development of self, I think another thing that makes it possible is the position of our eyes. We can only see in < 180 degrees, our eyes open & shut, and we happen to fall asleep. If the human subject was allowed free movement, the captor would want us to respond to commands .. we would need to be called .. differentiated amongst ourselves. Human Unit 1 is different from Human Unit 2. The captor calls the subject, and the subject chooses to respond, somehow. In the presence of punishment, the subject decides to carry out assigned duties to avoid punishment, or to gain reward: self-interest. Self.
Actually, it can be done. Depersonalization is relatively common: http://en.wikipedia.org/wiki/Depersonalization
However, do depersonalized people feel like they're no longer individuals, or do they feel like they no longer have control?
on another not to parent posters..... whether or not our sense of self is ongoing and inbreakable is certainly not decided or anywhere near scientific fact(yeah yeah, science only disproves, you know what i mean)...
all we can say is that we feel as if we have this continuity..... and that it apprars as if others feel the same way. memory is not like tape.... your sense of timing and events changes constantly, and what you perceive to be your unbroken, continuous sense of self and its memories is, in fact, almost universally incorrect on all kindsof things as it changes over time.... butyou (and i) will feel everything is in order. it is likely constantly teconstructed ad a survival trait.
Also look at surgical anaesthesia.... some theories on this, and subjectively i can see it, while you are under, you arent asleep...... you, your sense of self is gone, totally shut down. no dreams. no sense of how much time had passed when you wake........ not like a regular sleep whenyou at least havesome idea. it always feels like instant teleportation from the surgical suite to the recovery room, even if many hours have passed. then there arethepeople who just never come back.
we are far away from understnding consciousness... which is cool. weve batelyscratched the surface. were just nowrealizing the brain has far more plasticity than we thought a few years ago.... its still a hugemystery.
now take general anaesthesia...... during surgeries i can recall, i was simply gone. that time was simply time i didnt exist.
> As long as you're going to recreate memories indistinguishable from reality, you'll further need to create sensory input indistinguishable from reality. At which point you've simulated the entire encounterable universe
Not really, after all you only need to simulate the universe at the human level of the perception, which means you can ignore most of the computational complexity of simulating the actual universe.
>Each component technology might be interesting for its own sake, but aside from the philosophical point, what's the use? The carbon based versions are plentiful, and the construction process...
The cabron based forms are not very durable (80 years ? c'mon), they break easily, and don't perform very well...
Actually it's hard to build machines to last that long (how many 1932 cars are still functional today?)
The real advantage of the machines is that you can switch them off, open them up, replace parts, re-assemble them and switch them on again. You can replace parts of carbon-based lifeforms, but they don't always switch on again and opening them up often results in them being eaten by other carbon-based lifeforms. It's not cheap to do this to old machines, but its fairly reliable.
It's extremely readable and you may be surprised at how often people entirely miss the point when discussing it.
We have a brain like Zach have, but instead of being put inside Douglas tank, we are showed (and sensed) a virtual world. It's a trap, you can't prove that it's not the case.
Douglas also, in your memory, puts some strange definition/notion: infinity. The space and time are both infinite. But does that really make sense? If time wasn't infinite (and began somewhere) then we would know the Douglas trap. He is blocking your knowledge at some point.
This has been asked for thousands of years and its a great question. I think the answer which has been posed by many great thinkers and artists is often condensed down to :
The only reality that matters is the one you are in right now. Any higher reality, which you cannot access, essentially does not exist in any meaningful way to you.
"A technologically mature “posthuman” civilization would have enormous computing power. Based on this empirical fact, the simulation argument shows that at least one of the following propositions is true:
1) The fraction of human-level civilizations that reach a posthuman stage is very close to zero;
2) The fraction of posthuman civilizations that are interested in running ancestor-simulations is very close to zero;
3) The fraction of all people with our kind of experiences that are living in a simulation is very close to one.
If (1) is true, then we will almost certainly go extinct before reaching posthumanity. If (2) is true, then there must be a strong convergence among the courses of advanced civilizations so that virtually none contains any relatively wealthy individuals who desire to run ancestor-simulations and are free to do so. If (3) is true, then we almost certainly live in a simulation. In the dark forest of our current ignorance, it seems sensible to apportion one’s credence roughly evenly between (1), (2), and (3). Unless we are now living in a simulation, our descendants will almost certainly never run an ancestor-simulation."
4) The fraction of post-human civilizations that can build a simulation of a human-level civilization is very close to zero.
Simulating a human brain is many orders of magnitude beyond the capability of present-day computers, and Moore's Law isn't likely to continue forever. It is my belief that all of humanity, or even a single human, is much too complicated to be simulated by any current or future computer.
(From Nick Bostrom's home page: http://www.nickbostrom.com/ )
Once they have switched on an ancestor-simulation, it's probably illegal to turn it off. So it would mean committing the hardware and electricity to the simulation forever. Whatever they gain from running the simulation, they would be reluctant to incur infinite cost in doing it.
... but then we'd get strange effects (you know, perhaps things like wave-particle duality, quantum measurement, ...)
With this kind of simulations you can obtain, in theory, "what if" universes where alternatives solutions and technologies can be created. If you can observe them and the simulation rate is faster than your own time, it can be a rewarding experiment.
About universe observing, this story is cool: http://qntm.org/responsibility
Either #3 is false (and there are no simulations) or #3 is true (and the vast majority of sentient beings live in simulations, which means those running the simulation are probably simulations themselves.)
Yes a computer can run a simulation of an alternate version of itself, but then the speed must necessarily decrease as you go down the hierarchy (as you will see if you try to run one vmware vm inside another.) Or the simulated universe must lag behind the one containing the more powerful (thanks to Moore's law) host computer.
Edit: unless you have an infinitely-powerful computer like in that story. But then you could argue that every theoretically-possible simulation must be being run by someone somewhere. Good for us even if our society never discovers the infinitely-powerful quantum computer (at least we don't need to worry about being switched off, even if our creators die or their universe collapses.)
Interesting point anyway, basically a spin on the brain in a vat problem.
I love, and I fancifully imagine that he would too, the idea of referring to Greg Egan as 'they'.
The reason for that is that one of the problems with the concept of the Turing Test, is it's subject to the intelligence of the tester. A 6 year old boy with no talent in logic is much more likely to think a chat bot is a human, than an experienced CS researcher.
The dumber your tester, the dumber the AI needs to be to fool it. If you write a tester who is dumb as a rock, than it's trivial to write an AI that can fool it.
Zach and Douglas are only going into a lengthy conversation because Douglas went through the trouble of making Zach smart, knowledgeable and mimicking many human behaviors. If he made Zach as smart as a fundamentalist religious zealot. Then he could have just said "you're an AI because I told you so" and Zach would agree. But then again, that wouldn't be slightly as fun.
http://en.wikipedia.org/wiki/A_Mind_Forever_Voyaging
It's an interactive fiction game where you play the role of a computer that has only just realised that it's a computer. From your point of view, you've been living a human life with real experiences and a family etc.
The game manual included a great little short story: http://gallery.guetech.org/amfv/amfv.html
What about something concrete, like the sensation of pain? I can feel it, and everybody else can probably agree that they feel it too, even if they can't confirm that others do. How would you go about reproducing this sensation in a computer?
It's not clear to me that human thought and experiences can be reproduced by any amount of computer logic and memory. That doesn't mean it's impossible, but I think this is an unresolved question.
While your hand, say, might be physically hurt (say, it caught fire), to your brain it's only information coming in that says "now you should feel pain".
You can simulate that in a computer.
I will proudly stand up for the rites of computers as citizens of this country when they exhibit significant signs of ability to choose their own course and have opinions.
The computers will be our children, they will colinate the galaxy, and if we are lucky we can subscribe to the experience streams.
I guess you could say "two different ways to shuttle information around", but even that isn't quite right in the biological case. The way information is stored and retrieved in our brains is quite different from the way it is stored in today's computer memory, as demonstrated by the recent article posted on HN talking about how accessing a memory in one's brain involves recreating it from scratch, modifying it in the process. But that's not to say that we can't create a computer that more accurately reproduces the mechanisms of our brains, or that doing so is necessary to create a conscious computer.
there is indeed still debateabout whether our nervous system is fundamentallyelectric..... its certainly part of it, and cambeused to influenceit, but there are other chemicals and steuctures at work as well, as you say.
the point of theturing test is really to say that if you cant tell whether you are talking to a man or machine, and you believe a man is sentient, the you must assumethe machineis as well.... we haveno other mechanism to decide this. we caneven debate how we canprove ourselves to. ese tient.
Unless of course you postulate that there will be only one AI, because all instances will be networked with each other.
To take a different tack, I don't see how having a physical body has anything at all to do with being an individual. Certainly in our case out physical bodies are part of what makes us individuals, but what's to say that the same AI instantiated twice with different random seeds each time wouldn't produce two distinct individuals?
Consciousness produces logic as a tool to use. Logic does not produce consciousness.
Computers, by definition, are pure logic and rules. You can use logic to mimic consciousness, but nothing more.
Considering we have no falsifiable theory about what does produce consciousness, I don't see how you could possibly claim this.
You can't prove that you are conscious, nor can anybody else. I can perceive my own consciousness but I can't rigorously explain it. We have no way of determining if this alleged "consciousness" is a spectrum or binary. We can't test which life-forms are conscious and which are not. We can observe behaviors in animals that seem to imply consciousness, but is a dog conscious, or an insect, or an amoeba? We don't know.
And furthermore, if you believe in evolution, you have to believe that there was no consciousness and then consciousness at some point was created where there was none before. If logic doesn't produce consciousness, what does produce it? And whatever produced it in cellular tissue millions of years ago, who's to say we couldn't likewise produce it in a die of silicon, which like our brains is highly electrical?
Actually, you don't http://en.wikipedia.org/wiki/Panpsychism
There are a lot of complex and emergent systems that demonstrate this property of following simple rules but exhibiting complex behaviors. Conway's game of life, rule 110 in basic cellular automata, how ants work together, fireflies blink in synchrony without global communication are all examples.
And while none of these demonstrate that computers can produce consciousness, they give a hint that computers may be able to do so, despite it following rules to the T.
If our brains can produce consciousness there is no physical reason we know of why a computer can't be conscious. How do you know if everyone around you is conscious? You just assume they are, because you are. In reality you have no way of telling.
The interesting thing about this post is that the robot believes it is human, because it believes it is conscious. If it acts in a way that we recognise as conscious and tell us it is, at this point we have just as much evidence for believing the robot is conscious as we do for every human.
What do Expertise, http://en.wikipedia.org/wiki/Flow_(psychology) and the advise to go for a walk or sleep on a problem have (or downplay) in common?
http://en.wikipedia.org/wiki/Benjamin_Libet#Implications_of_...
And AFAIK, we don't even know if it's computable.
I find your superstition disturbing.
This statement is not proved (and maybe not provable at all, not to mention probably untrue).
Also, have you considered the possibility a THIRD factor besides logic/consciousness (like information complexity, feedback loops etc) to produce BOTH consciousness and logic?
Computers, by definition, are pure logic and rules. You can use logic to mimic consciousness, but nothing more.
Who proved that, when and how? This is just a circular (vicious) argument, especially as denoted by the "by definition" part.
Second, "just two different ways to shuttle electrons around" would hold true if information exchange was the only thing that mattered. Which could be true, but what if the actual chemical etc substances used are also important? Like how you can model audio (say, a WAV file), but you cannot model actual sound.