My problem with time travel and teleportation

[Lord Joeltion]

It's not 'it only matters if its visible to other people', its 'it only matters if it has a causal consequence to events which occur in the universe'. If e.g. I were somehow able to do something that 'should' swap your self-awareness with the self-awareness of someone else, but when I did so nothing at all changed - neither you nor the other person reported feeling like you jumped across the room - then I would conclude that in fact self-awareness (at least, as defined by this mysterious power to swap them) does not exist. There could be some other property of self-awareness that exists, but it would have to be different than whatever thing I thought I was swapping was.

The very fact that you used "mysteriously" to describe the process by which an AI is successfully transported without creating a copy, should make you realize the case at hand: We don't frigging know what consciousness is or how it works or why something becomes spontaneously self aware at some point between the moment of conception and the moment a baby starts to react to stimulation.

Why is it so hard to swallow, my friend? Is "I don't know" to hard to accept? Does science really needs to have all the answer? Guess what, it never will. And this is coming from a guy who strongly believes that science is the only way Humanity has to evolve into something better; the only torch we hold against a bleak, cold, deadly universe.

Anyway, this phrasing simply doesn't make sense:

"(if) I were somehow able to do something that 'should' swap your self-awareness with the self-awareness of someone else, but when I did so nothing at all changed - neither you nor the other person reported feeling like you jumped across the room - then I would conclude that in fact self-awareness (at least, as defined by this mysterious power to swap them) does not exist."

Your premise is that you have a certain device that should accomplish certain task. When said task isn't accomplished your conclusion is... that the task is impossible to perform? That is like saying that a soup is broken because your pencil doesn't allow you to eat it. Your logic there is completely erroneus.

If a device can't accomplish the task you thought it was meant to perform, it's either broken or useless. The device doesn't work, not the phenomena.

If we posit that the teleporter is a device that by definition doesn't preserve some property 'self-awareness', it's the same story. If we teleport someone, and nothing at all changes, then the property 'self-awareness' as defined by the statement 'this is a property not preserved under teleportation' doesn't exist - that is to say, it has no consequence, not even to you (because if it had consequence to you, you'd behave differently as a result of that consequence - but in the scenario with the teleporter, you didn't behave differently).

Problem being: A doppleganger is indistinguishable from me, if he is in fact a perfect copy. Yet, in reality, he isn't me. Your argument would work if my position was that "A teleporter is unable to transport the property of self-awareness of the subjects it transmit through space". But that is absolutely not my premise in the least.

My real argument is: "Since there is no way to prove that somebody is somebody and not a copy of themselves; a Teleporter that dematerializes a subject should not be relied upon for personal transport". Either the teleporter is a "real" teleporter (something that should involve magic, for our current understanding of how the universe works); or it is a Murder-Cloning Machine. You can't prove either. You will never know. Unless you try it. There is certain risk involved. If you can't understand why, then I have no simple way to explain how our current understangding of the universe works. It would be too much of a hassle :P

Well, you do this every day when you go out in society and people are driving/biking/etc around you - at any time, their mistake could kill you. Whenever you get in a car or an airplane you're trusting the engineers (or at the very least, trusting the track record of the technology). When you take medicine from the doctor, of course it could be botched or defective or something, or maybe something you're allergic to, or the prescription could be filled incorrectly. In the end, everything you do involves a risk. Even doing nothing involves a risk - that's one day older and one day more wear and tear on your body.

Sigh.... Lemme put it simply: Being rational isn't about taking zero-risks (that is in fact, irrationality). Being rational is weighing outcomes and making everything in your power to avoid unnecessary risks. The most rational thing to do, when you have even the tiniest suspicion of stepping inside a Suicide Machine is, well... not to step on it. Then again, some people like playing Russian Roulette. I don't judge those people. I simply don't play.

Forget about the tinkering for a moment and lets talk about the backup. Every backup you have that you're willing to consider yourself squares your probability of dying in a given interval (keep in mind that p^2 < p, since p<1). If making a backup involves some risk of death to the source material - lets say its a 30% risk of death each time - you're still going to be ahead if you spend all your time constantly making backups. Sure you have that first 30% death risk, but then it becomes effectively 10% for the next backup, then 3%, then 1% - but so too do all your other risks of death shrink. So eventually, Robo-Joeliton gets the option of either choosing guaranteed death eventually, or a coin flip with a roughly 45% chance of dying immediately and a 55% chance of ironclad immortality.

I don't want to forget the tinkering part. Don't tinker with me. Don't teleport me against my will!

I congratulate you for doing the math there, but I would lie if I said I paid much atention to it. Numbers have little meaning in the face of complete uncertainty.

I'd take that coin flip. I'm not going to say something like 'its the only rational move' because e.g. other, less risky backup technologies might come into existence if you were to wait.

A coin flip involving "death" vs. "immortality" is VERY, VERY, VERY, VEEEEEEEERY different from one that involves "getting somewhere faster" vs. "being replaced by a doppleganger". If you can't tell the difference then I have nothing else to tell you. But I admit, this conversation was fun

As I've said elsewhere in the thread, yes, we have already discovered how to make personal experience observable to a third party in an objective fashion. There's loads of research on this.(snip snip snip snip)

Uh, what? Yes, there is lots of research; but no, we are far to having definitive answers (or even a proper method to discern what is the key to real intelligence). If we had the answers, then probably 99% of the problems Linguists, Psychologists, AI researchers and Phylosophers are currently trying to solve, would have been already solved. I know we are doing hell of improvement and all... but we haven't scratched the surface yet. At least that is my opinion and the opinion of most scientists I am aware of*

*Not invoking any kind of authority; I'm just trying to clarify this isn't just my personal belief, or something pulled from my "back-ups".

So, in summary: reincarnation (the closest thing to the resurrection of the example, because the ability to resurrect D&D characters is not that relevant to hypothetical real world transporters) could theoretically be real given about a dozen natural forces that haven't been considered in modern physics yet (fine, I'll accept that, we don't know nearly everything yet, as long as we still find new cat sized land mammals we can't possibly claim to know stuff like this for sure), and the thing that most people would call a soul that actually reincarnates into the next body and that carries my personality is me and transports over just fine, but there could be a second spiritual thing that actually handles the transition of that me into a new body after I die, and it's possible that a transporter decouples that separate thing from me, because a thing in charge of taking me to a new body can't handle me being transported to a new body.

I used D&D to avoid breaking the rules. My point was to present you a set of beliefs (which may or may not reflect RL beliefs) that would make your test innacurate/inconclusive. It wasn't about how somebody could "move the post goals" on the subject of souls; it was about how false positives are still very much possible, depending on the actual definition of "soul" you attempt to work on (which, again, it's far from having any kind of consensus).

Anyway, I wasn't disagreeing with your proposed experiment. I very much agree with your other conclusions, in fact. Where I disagree is that it can provide a definitive answer, like the one you claimed. It's certainly the best approach, but it's still far from conclusive.

At this point, honestly, that seems reaching. I'll politely listen to anyone who has concerns over a spiritual part of my personality that might get lost, even though it's never been my personal favorite explanation for my behavior, but ones the argument gets to this point, I'll take my chances with the transporter.

Oh, yeah. I agree, I think. As mentioned before, my concern isn't spiritual, it's more like psychological. But when put between a rock and a hard place, I see no reason why risking traveling through a teleport isn't worth the risk. But it should be one heck of a hurry I'm having, in order to justify it myself.

Which Robert Angier was the real Robert Angier???

The one who killed Barakapool for good, of course.

[NichG]

The very fact that you used "mysteriously" to describe the process by which an AI is successfully transported without creating a copy, should make you realize the case at hand: We don't frigging know what consciousness is or how it works or why something becomes spontaneously self aware at some point between the moment of conception and the moment a baby starts to react to stimulation.

Why is it so hard to swallow, my friend? Is "I don't know" to hard to accept? Does science really needs to have all the answer? Guess what, it never will. And this is coming from a guy who strongly believes that science is the only way Humanity has to evolve into something better; the only torch we hold against a bleak, cold, deadly universe.

Maybe this is a point I should have gotten to earlier, but we don't actually need definite answers or explanatory models before finding and doing things that work. You can test things empirically.

For example, one level of understanding of medicine is 'when patients with certain symptoms consume this compound that I harvested from a particular kind of mold, their survival rate is increased by 15%'. That doesn't mean I know how that compound is synthesized chemically by the organism, what its mode of action is within the body, how the symptoms come about and what their underlying cause is and so on. Of course, we then study things further to develop that level of understanding, but its not a necessary pre-condition before we can assess the question 'is this compound useful as medicine?'

Similarly, lets say I don't know what the neurological basis of the ability to see color is. One way to find out more is to perform ablation studies. That is to say, I have a bunch of rats and in each rat I destroy some part of their body - their left leg, one of their lungs, bits and pieces of their brain, etc. In doing so, I can't distinguish between whether its the heart or the brain that's responsible for color vision (because if I destroy either, the ability is lost), but I can definitively say 'it has nothing to do with their left leg' because I can create cases where color vision remains even when their leg is destroyed.

The teleporter is essentially an ablation study for the soul. We could imagine a sequence of better and better teleporters - the worst one takes a photo of you and then shapes a puddle of undifferentiated biomass into the shape of your body. The next one puts together a generic human body from a template, but tries to match your MRI scan in terms of proportions. The next one takes chemical samples and tries to match levels of neurotransmitters, etc. The next one maps the pattern of synapses, but grows new neurons from scratch for you. The next one copies things down to the macro-molecular level, but doesn't copy e.g. the arrangement of water molecules. The next one puts every atom in its place precisely.

At some point in that sequence, you'll go from 'there's now a corpse on the teleporter pad' to 'there's something that manages to live'. And at a different point you'll go from 'there's something that manages to live' to 'there's something that exhibits cognitive function'. And at some point (in the universe of the thought experiment of this thread) you'll get 'no one can tell the difference between the copy and the original, including close friends and family'.

So the level at which each of those properties is preserved tells you about the necessary and sufficient conditions for that property being copied, and gives you empirical evidence about the fidelity of the copy.

If thousands of people go through the teleporter every day, and still there's no actual reported teleporter-related diseases or side-effects, then the empirical evidence is that there isn't anything important that's being missed. That evidence would stand even if we don't have an explanatory story of why.

"(if) I were somehow able to do something that 'should' swap your self-awareness with the self-awareness of someone else, but when I did so nothing at all changed - neither you nor the other person reported feeling like you jumped across the room - then I would conclude that in fact self-awareness (at least, as defined by this mysterious power to swap them) does not exist."

Your premise is that you have a certain device that should accomplish certain task. When said task isn't accomplished your conclusion is... that the task is impossible to perform? That is like saying that a soup is broken because your pencil doesn't allow you to eat it. Your logic there is completely erroneus.

If a device can't accomplish the task you thought it was meant to perform, it's either broken or useless. The device doesn't work, not the phenomena.

Here the 'should accomplish' of the teleporter is 'destroying the self-awareness of the original'. It's been argued in this thread that, philosophically, a teleporter that destroys the original body and builds a new body must destroy the self-awareness of the original. However, when the copy exhibits all evidence of having a continuity of experience with the original (and the original being left alive also exhibits the same evidence), the conclusion must be that this term 'self-awareness' as defined from the philosophical standpoint isn't actually referring to the thing you're experiencing when you experience 'self-awareness'.

Problem being: A doppleganger is indistinguishable from me, if he is in fact a perfect copy. Yet, in reality, he isn't me. Your argument would work if my position was that "A teleporter is unable to transport the property of self-awareness of the subjects it transmit through space". But that is absolutely not my premise in the least.

My real argument is: "Since there is no way to prove that somebody is somebody and not a copy of themselves; a Teleporter that dematerializes a subject should not be relied upon for personal transport". Either the teleporter is a "real" teleporter (something that should involve magic, for our current understanding of how the universe works); or it is a Murder-Cloning Machine. You can't prove either. You will never know. Unless you try it. There is certain risk involved. If you can't understand why, then I have no simple way to explain how our current understangding of the universe works. It would be too much of a hassle :P

If you can't possibly know unless you try it, then necessarily you can't possibly know even if you try it either.

Uh, what? Yes, there is lots of research; but no, we are far to having definitive answers (or even a proper method to discern what is the key to real intelligence). If we had the answers, then probably 99% of the problems Lunguists, Psychologists, AI researchers and Phylosophers are currently trying to solve, would have been already solved. I know we are doing hell of improvement and all... but we haven't scratched the surface yet. At least that is my opinion and the opinion of most scientists I am aware of*

Okay, this is going to be a long aside, so spoilering it to not make the thread even more confusing.

Spoiler: Stuff about current AI research

Show

As an AI researcher, the detailed study of how human cognition at a microscopic works is basically irrelevant to the actual advances in AI research we've seen in the last few years. Things like physiologically realistic spiking neuron models tend to be really bad at learning compared to what you can build on a computational substrate taking into account that you have the advantages of that substrate to work with which biology doesn't. From the point of view of AI, human-level cognitive ability in perceptual tasks is generally achievable using standard techniques and sufficient data, no magic needed. Similarly, its not hard to exceed human-level performance in inference tasks - aggregating multiple sources of evidence towards a set of questions. Transfer learning, one-shot generalization, and things like that went from 'mostly miss' to 'some hits' in the last year or so - there's a trick involving the intersection of memory and attention which was invented in parallel in something like 10-20 papers, and it makes the difference between needed 10000 examples to learn and just needing 10 examples to learn, with a caveat that you do need a pretty diverse set of past experiences in weakly-related cases. Facebook's scary-good face recognition stuff is an example of this kind of technique - even the publically available version can identify a person based on only one or two photos, with a 1/1000 false positive rate and a 1/200 false negative rate.

On the other hand, control is still hard - human-level performance is hit or miss. Motivation and goal generation are currently mostly miss - we're still at the level of handmade motivation functions like Curiosity and Empowerment rather than learning to set its own goals (with a few exceptions), and the current direction of investigation is IMO overly obsessed with 'reward functions' so it's kinda stuck there. In these problems at least, I do think there's some fundamental change of perspective that the field needs. Meta-optimization is kind of hovering around the fringes here (so called recursive self-improvement) and has had some successes, but at a very low payoff per increase of algorithmic complexity and very bad diminishing returns so they're not really used much.

Language generation is hit or miss - we can hit human-level performance in narrow linguistic tasks (e.g. negotiation) but not broad ones. Symbolic manipulation and logical reasoning is still hard - we're solidly below human-level performance there. The narrow stuff is very impressive though - one recent paper lets you take a screenshot of a UI and the AI writes code that reproduces the same UI, for example. There's another paper where you sketch a figure and the AI can write a short pieces of LaTeX code that not only generates the figure, but can extrapolate it to different sizes (e.g. you draw a small graph of a tree, and it can make a larger graph using that as a seed). In general, extrapolation, program induction, and symbolic manipulation seem to be one research 'set' together with each-other.

But for all of this, I don't think consciousness or self-awareness is particularly a 'problem' for AI - that is to say, I can't point to any particular AI task where the logical flow of solving the task would require 'first we have to solve self-awareness, then we can do X'. It may well be that architectures which solve a given task will also incidentally 'solve' self-awareness, or that self-awareness in AI is more of a characterization problem (it might have it or not, but are we even looking for it?). If anything, my bet here would be that if we were somehow able to agree on what to use the term for, we'd already find significant self-awareness in recurrent attentional models - they do things like fuse external sensory streams with internally generated sensory streams, recursively attend to their own attention mechanism, etc, so you have all of the colloquial phenomena associated with self-awareness such as perception of your own thinking process, ability to imagine things, etc. AI of this sort is good at games like Sokoban for example, which require visualizing future sequences.

Anyhow, that's for AI at least. Neuroscience, linguistics, and philosophy all have their own questions and problems of course.

[Tyndmyr]

For example, if A is my past self and B is my current self, then:

A => B
B not => A

If => is intended to mean is the same as, you ought to use the equal sign. And, seriously, this is a fundamental property of mathematics, logic and reality you're arguing with. If A = B, then B = A.

If you mean that B is a copy of A, well, no argument there.

But you cannot actually claim that B *is* A, and yet A *is not* B. If they are the same entity, then both statements must necessarily be true.

Either that, or feel free to disprove all of known mathematics.

[georgie_leech]

It requires experimental validation - namely, we have to go and check to see that the two objects (and the process of doubling the construction) behave in the correct way. For example, if I were to sneak in and make a nick in one of the lego pieces in your first object, would I find that same nick in the second object? If I were to embed a microchip inside the plastic, would I find that you managed to copy all the details of the microchip too?

If not, then it's not the same situation as the A/D objects.

Yes, that is what I was getting at; that these are two clearly distinct objects, such that even if no test could distinguish between them, you still clearly have two of them. Do you disagree?

[NichG]

If => is intended to mean is the same as, you ought to use the equal sign. And, seriously, this is a fundamental property of mathematics, logic and reality you're arguing with. If A = B, then B = A.

It's not intended to mean 'the same as', it's intended to mean this idem property that we've been talking about, where if x => y then making a change to x also implies a change occurs to y.

The mathematical basis that I've been using for the => operator is causality, specifically in the form of Pearl's do-calculus, which formulates causality by positing interventions and then tracks how the consequences of the interventions propagate. Unlike correlation, causality is directional. A causes B doesn't imply B causes A. The operator x => y means something like 'y is entirely caused by x'.

But you cannot actually claim that B *is* A, and yet A *is not* B. If they are the same entity, then both statements must necessarily be true.

Are you the same entity as your past self? Obviously you aren't 'equal' to your past self.

Yes, that is what I was getting at; that these are two clearly distinct objects, such that even if no test could distinguish between them, you still clearly have two of them. Do you disagree?

Again we have to be very careful about time ordering here. They are now two clearly distinct objects. They were one object.

[georgie_leech]

Again we have to be very careful about time ordering here. They are now two clearly distinct objects. They were one object.

In what way? They have the same properties, but that's because the second block of Lego blocks was built to the same specifications with the same types and colours of pieces; I specifically said "construct" to mean that there wasn't some magical copying process, but careful assembly instead.

[NichG]

In what way? They have the same properties, but that's because the second block of Lego blocks was built to the same specifications with the same types and colours of pieces; I specifically said "construct" to mean that there wasn't some magical copying process, but careful assembly instead.

Ah okay, so, before you asked me whether or not the constructed lego object had the same relationship to the source one as the teleported object did to the source one, and I replied that it depends on whether the 'specification' also includes duplicating everything about the source lego object to the extent that there is no possible single thing you could do to both objects that would get them to behave differently. If, for example, the careful construction process doesn't involve copying the scratch marks/etc, then e.g. 'looking closely' is a thing you can do to distinguish them.

So yeah, in this case the objects at least were not one object becoming two, I agree. (If we talk about 'the layout' though, its one layout becoming two, but to talk about that carefully we need to introduce new things so maybe not yet?)

[Lord Joeltion]

Maybe this is a point I should have gotten to earlier, but we don't actually need definite answers or explanatory models before finding and doing things that work. You can test things empirically.

You certainly can build a bomb without knowing chemistry. That doesn't mean your expertise on bombs should be trusted. The empirical method falls short when a specific level of certainty is required/demanded.

In any case, this thought experiment would be useful if you already have a definitive answer for all its relevant propositions. If you propose a machine that works under undefined properties, you can't seriously expect me to give you a meaningful answer. If you want an answer from me other than "it's unreliable" you should provide me facts, not "it works because it works".

For example, one level of understanding of medicine is 'when patients with certain symptoms consume this compound that I harvested from a particular kind of mold, their survival rate is increased by 15%'. That doesn't mean I know how that compound is synthesized chemically by the organism, what its mode of action is within the body, how the symptoms come about and what their underlying cause is and so on. Of course, we then study things further to develop that level of understanding, but its not a necessary pre-condition before we can assess the question 'is this compound useful as medicine?'

That is an issue intrinsic to the medical field. The body is way too chaotic and individuals vary wildly for medical research to work under any kind of certainty. That very same complexity makes them use the empirical method; not because it is the best theoretical model, but because it's the only they can work with. And that's precisely the reason why medicines have so many contraindications. They aren't custom made. They can't provide certainty, only some degree of safety (as in, "at least it won't kill you").

On the other hand, engineers don't go around blindly firing rockets to the sky until SOMEHOW an Apollo 30 reaches Mars. They use the appropriate method, which involves a lot more theory than plain empiricism, and is infinitely more reliable.

Similarly, lets say I don't know what the neurological basis of the ability to see color is.

You don't know what the neurological basis of the ability to anything is. The most scientist have reached is getting to know what areas of the brain are specialized for some kind of interaction with the environment; and little more than that. I really don't get what was your point with hurting theoretical rodents. That's not how we discovered how vision works, or the structure of the eye-brain interaction.

The teleporter is essentially an ablation study for the soul. We could imagine a sequence of better and better teleporters - the worst one takes a photo of you and then shapes a puddle of undifferentiated biomass into the shape of your body. The next one puts together a generic human body from a template, but tries to match your MRI scan in terms of proportions. The next one takes chemical samples and tries to match levels of neurotransmitters, etc. The next one maps the pattern of synapses, but grows new neurons from scratch for you. The next one copies things down to the macro-molecular level, but doesn't copy e.g. the arrangement of water molecules. The next one puts every atom in its place precisely.

At some point in that sequence, you'll go from 'there's now a corpse on the teleporter pad' to 'there's something that manages to live'. And at a different point you'll go from 'there's something that manages to live' to 'there's something that exhibits cognitive function'. And at some point (in the universe of the thought experiment of this thread) you'll get 'no one can tell the difference between the copy and the original, including close friends and family'.

Mmm... And I though *I* was very creative. Sorry, go on...

So the level at which each of those properties is preserved tells you about the necessary and sufficient conditions for that property being copied, and gives you empirical evidence about the fidelity of the copy.

It's a copy. That's the problem here. You are trying to convince me that the consciousness in *me* will be magically transported along my body to the *copy* of me. You are trying to convince me about something without any scientific background or real proof about what makes people aware of themselves. You know how that sounds to me? It sounds very similar to something that is against the rules mentioning.

You won't convince me without the proper science (which involves a LOT more than a machine that "appears" to work as intended), pal. It would be silly of me to accept your "empirical proof" (the weakest form of proof in science) or accept your beliefs just like that.

If thousands of people go through the teleporter every day, and still there's no actual reported teleporter-related diseases or side-effects, then the empirical evidence is that there isn't anything important that's being missed. That evidence would stand even if we don't have an explanatory story of why.

I think you have NOT watched the video. Have you watched the video(s) at all? I think it would save me from repeating yet again why, no matter how much you rephrase it; the very fact that you have no way to explain how the machine works, makes it absolutely unsafe for my consciousness. "Empirical demonstration" is useless, because it doesn't provide any answer at all.

Here the 'should accomplish' of the teleporter is 'destroying the self-awareness of the original'. It's been argued in this thread that, philosophically, a teleporter that destroys the original body and builds a new body must destroy the self-awareness of the original. However, when the copy exhibits all evidence of having a continuity of experience with the original (and the original being left alive also exhibits the same evidence), the conclusion must be that this term 'self-awareness' as defined from the philosophical standpoint isn't actually referring to the thing you're experiencing when you experience 'self-awareness'.

Your conclusion is simply bogus. Just because you can't notice the difference it doesn't mean there was never a difference to begin with. It's a lack of understanding on your side, it says little about reality. Your rationalization of the situation is akin to somebody claiming that because no crime can be demonstrated; no crime was ever committed. Reality doesn't care what think about evidence. There was a time where we thought dinosaurs had reptilian skin because that was what we understood from the evidence available. Guess how Jurassic Park should have looked like, were it more "realistic". So no, until science actually demonstrates something, intuition and logic will always trump whatever empiricism tries to argue.

If you can't possibly know unless you try it, then necessarily you can't possibly know even if you try it either.

I'm not sure if you tried to sound smart or not, but I don't see what you actually meant with this. I don't know what number will come from a dice unless I roll it (all I know is the range, and the probability from each outcome; yet each different outcome shares the same probability). Does that mean even when I roll it AND see it I still don't know what number did fall on?

Okay, this is going to be a long aside, so spoilering it to not make the thread even more confusing.

Spoiler: Stuff about current AI research

Show

As an AI researcher, the detailed study of how human cognition at a microscopic works is basically irrelevant to the actual advances in AI research we've seen in the last few years. Things like physiologically realistic spiking neuron models tend to be really bad at learning compared to what you can build on a computational substrate taking into account that you have the advantages of that substrate to work with which biology doesn't. From the point of view of AI, human-level cognitive ability in perceptual tasks is generally achievable using standard techniques and sufficient data, no magic needed. Similarly, its not hard to exceed human-level performance in inference tasks - aggregating multiple sources of evidence towards a set of questions. Transfer learning, one-shot generalization, and things like that went from 'mostly miss' to 'some hits' in the last year or so - there's a trick involving the intersection of memory and attention which was invented in parallel in something like 10-20 papers, and it makes the difference between needed 10000 examples to learn and just needing 10 examples to learn, with a caveat that you do need a pretty diverse set of past experiences in weakly-related cases. Facebook's scary-good face recognition stuff is an example of this kind of technique - even the publically available version can identify a person based on only one or two photos, with a 1/1000 false positive rate and a 1/200 false negative rate.

On the other hand, control is still hard - human-level performance is hit or miss. Motivation and goal generation are currently mostly miss - we're still at the level of handmade motivation functions like Curiosity and Empowerment rather than learning to set its own goals (with a few exceptions), and the current direction of investigation is IMO overly obsessed with 'reward functions' so it's kinda stuck there. In these problems at least, I do think there's some fundamental change of perspective that the field needs. Meta-optimization is kind of hovering around the fringes here (so called recursive self-improvement) and has had some successes, but at a very low payoff per increase of algorithmic complexity and very bad diminishing returns so they're not really used much.

Language generation is hit or miss - we can hit human-level performance in narrow linguistic tasks (e.g. negotiation) but not broad ones. Symbolic manipulation and logical reasoning is still hard - we're solidly below human-level performance there. The narrow stuff is very impressive though - one recent paper lets you take a screenshot of a UI and the AI writes code that reproduces the same UI, for example. There's another paper where you sketch a figure and the AI can write a short pieces of LaTeX code that not only generates the figure, but can extrapolate it to different sizes (e.g. you draw a small graph of a tree, and it can make a larger graph using that as a seed). In general, extrapolation, program induction, and symbolic manipulation seem to be one research 'set' together with each-other.

But for all of this, I don't think consciousness or self-awareness is particularly a 'problem' for AI - that is to say, I can't point to any particular AI task where the logical flow of solving the task would require 'first we have to solve self-awareness, then we can do X'. It may well be that architectures which solve a given task will also incidentally 'solve' self-awareness, or that self-awareness in AI is more of a characterization problem (it might have it or not, but are we even looking for it?). If anything, my bet here would be that if we were somehow able to agree on what to use the term for, we'd already find significant self-awareness in recurrent attentional models - they do things like fuse external sensory streams with internally generated sensory streams, recursively attend to their own attention mechanism, etc, so you have all of the colloquial phenomena associated with self-awareness such as perception of your own thinking process, ability to imagine things, etc. AI of this sort is good at games like Sokoban for example, which require visualizing future sequences.

Anyhow, that's for AI at least. Neuroscience, linguistics, and philosophy all have their own questions and problems of course.

It must be cool being you! I have great respect for people who work in your field, honestly. Anyway, I was speaking broadly of the fields that are related in any way to Cognitive Science. Obviously I was exaggerating with the relevance of self-awareness in particular; but I was referring to the fact that as CS is concerned, the day they solved the mysteries of SA, it means they have already solved every other major problem. It would also be considered a major breakthrough for most of those fields, mostly for the paradigmatic change it could entail.

The aditional info on AI is appreciated

[Bohandas]

That is an issue intrinsic to the medical field. The body is way too chaotic and individuals vary wildly for medical research to work under any kind of certainty. That very same complexity makes them use the empirical method; not because it is the best theoretical model, but because it's the only they can work with. And that's precisely the reason why medicines have so many contraindications. They aren't custom made. They can't provide certainty, only some degree of safety (as in, "at least it won't kill you").

On the other hand, engineers don't go around blindly firing rockets to the sky until SOMEHOW an Apollo 30 reaches Mars. They use the appropriate method, which involves a lot more theory than plain empiricism, and is infinitely more reliable.

What about nuclear physics? They had to shoot ten quintillion calcium atoms at a reaction target for just three of them to hit and form oganesson, and it took four months to do.

[Lord Joeltion]

What about nuclear physics? They had to shoot ten quintillion calcium atoms at a reaction target for just three of them to hit and form oganesson, and it took four months to do.

What about it? I don't see how verifying a theoretical particle (or atom, in this case) is related to the methods of applied sciences such as medicine. They had a pretty fair knowledge of what they were doing and the potential results. It wouldn't have taken four months if we had the technology to actually manipulate subatomic particles to arrange them as we desire. It was a technical impairment, not a theoretical one.

[NichG]

@joeltion: I think a line by line response is going to be extremely confusing and disjointed, so trying to condense things here...

The point about the ablation studies is that it's possible to set up situations where you can know that something cannot be true, without knowing why it cannot be true. That's actually stronger than a first principles theoretical result, because if the theory said otherwise it would be disproven.

In the end, the result of the experiment is the result of the experiment and theory, logic, and philosophy have to be consistent with it. The reason is that any kind of theoretical basis for something has assumptions and axioms underpinning it, which the theory itself has no way of proving or disproving. So empirical evidence is used as a way to select those things which have axioms consistent with reality from those things which have axioms inconsistent with reality.

What I'm claiming is that if you say 'I have this thing which I'll call consciousness and I experience having it' and then 'I have a theoretical or philosophical deduction that a teleporter making a copy of me should not copy my consciousness' then if you observe the fact 'people go through the teleporter and do not report any disruption in consciousness' then something is wrong with your theoretical/philosophical deduction. It's inconsistent.

The reason its inconsistent is the italicized part specifically. If you said 'maybe we have this thing 'consciousness' that we can't experience, and the teleporter wouldn't copy it, but it might matter' then you'd be consistent (but questionable due to introducing new properties without any justification). But specifically because you claim to experience your consciousness, you're saying that a person can actually measure their own consciousness and exhibit the result of that measurement by e.g. talking about how they feel they're conscious.

So with my comment 'you necessarily wouldn't know if you tried it either', this is what I'm getting at. If you claim that your verbal report about your experience of consciousness consists of positive evidence for the existence and properties of consciousness, you have to allow in general that verbal reports about the experience of consciousness are positive evidence for the existence and properties of consciousness. You don't get to say 'listen to me now, but don't listen to the guy who comes out of the teleporter'.

[georgie_leech]

Ah okay, so, before you asked me whether or not the constructed lego object had the same relationship to the source one as the teleported object did to the source one, and I replied that it depends on whether the 'specification' also includes duplicating everything about the source lego object to the extent that there is no possible single thing you could do to both objects that would get them to behave differently. If, for example, the careful construction process doesn't involve copying the scratch marks/etc, then e.g. 'looking closely' is a thing you can do to distinguish them.

So yeah, in this case the objects at least were not one object becoming two, I agree. (If we talk about 'the layout' though, its one layout becoming two, but to talk about that carefully we need to introduce new things so maybe not yet?)

What I mean is that if you scratched the first Lego thing in some way before you started making the second, the second would have the scratch mark. If you for some reason snuck (apparently my spellcheck doesn't recognise that as a word) a microchip into it before hand, you would include that as wel because you're carefully analyzing the thing before you start on the second one. But if you did either of those while the second Lego thing was being made, such changes wouldn't be carried over. Does that make the scenario a little more clear?

[NichG]

What I mean is that if you scratched the first Lego thing in some way before you started making the second, the second would have the scratch mark. If you for some reason snuck (apparently my spellcheck doesn't recognise that as a word) a microchip into it before hand, you would include that as wel because you're carefully analyzing the thing before you start on the second one. But if you did either of those while the second Lego thing was being made, such changes wouldn't be carried over. Does that make the scenario a little more clear?

Yes, that's clear. In that case I'd say it's one object that split and became two objects (via an unusual intermediary perhaps, but that doesn't actually matter because the intermediary is perfect at its job).

[Lord Torath]

... snuck (apparently my spellcheck doesn't recognize that as a word)...

Try "sneaked".

[georgie_leech]

Yes, that's clear. In that case I'd say it's one object that split and became two objects (via an unusual intermediary perhaps, but that doesn't actually matter because the intermediary is perfect at its job).

So would you view either of the objects as the original, even if that's not a physical quality?

[Bohandas]

What about it? I don't see how verifying a theoretical particle (or atom, in this case) is related to the methods of applied sciences such as medicine. They had a pretty fair knowledge of what they were doing and the potential results. It wouldn't have taken four months if we had the technology to actually manipulate subatomic particles to arrange them as we desire. It was a technical impairment, not a theoretical one.

If we could manipulate things with subatomic precision medicine wouldn't be so fiddly either

[The Ari-tificer]

A related question: When a cell copies itself through mitosis, which of the resulting cells is the original and which is the copy?

When cells go through mitosis, they actually split in two. So the answer to your question is simultaneously both and neither.

[NichG]

So would you view either of the objects as the original, even if that's not a physical quality?

Yes, that makes the most sense to me.

[Tyndmyr]

It's not intended to mean 'the same as', it's intended to mean this idem property that we've been talking about, where if x => y then making a change to x also implies a change occurs to y.

The mathematical basis that I've been using for the => operator is causality, specifically in the form of Pearl's do-calculus, which formulates causality by positing interventions and then tracks how the consequences of the interventions propagate. Unlike correlation, causality is directional. A causes B doesn't imply B causes A. The operator x => y means something like 'y is entirely caused by x'.



Are you the same entity as your past self? Obviously you aren't 'equal' to your past self.



Again we have to be very careful about time ordering here. They are now two clearly distinct objects. They were one object.

Generally, people are indeed treated as the same person as their past selves. That said, there's also implict acknowledgement that humans change over time.

If your term of "idem" means merely "comes from", then sure, a copy comes from you. But it's not you, in the same way that you are not your parents, despite your parents having caused you.

Identity does not stem from causality.

And, getting back to my first line, it is clear that identity does not stem solely from sameness. Everyone acknowledges that a child changes in becoming an adult, but has no issue describing them as the same individuality. Therefore, constructing another object that is similar to an existing object cannot actually *be* that object.

Yes, that's clear. In that case I'd say it's one object that split and became two objects (via an unusual intermediary perhaps, but that doesn't actually matter because the intermediary is perfect at its job).

The word "split" has a specific meaning that does not appear to apply here. I concur that the second lego construction would be very similar to the first, but I see no reason why we should refer to the process of creating as "splitting" merely because the copy is a good one.

[NichG]

Generally, people are indeed treated as the same person as their past selves. That said, there's also implict acknowledgement that humans change over time.

If you have two things which have the exact same mathematical relationship as an object has with its past self, what's your basis for considering one to be the same and the other different?

[georgie_leech]

Yes, that makes the most sense to me.

Then that's where we disagree. If the first Lego thing is in my hand while the second is being constructed, intuitively, the first is the original, as there's no continuity with the second. That is, it's accurate enough to say that the second "comes from" the first, but there remains separation; there were steps in between the existence of the second Lego thing and the first, while the first necessarily existed at the same time as itself, so the first is the original.

[Tyndmyr]

If you have two things which have the exact same mathematical relationship as an object has with its past self, what's your basis for considering one to be the same and the other different?

They don't, though.

Look at the two lego structures. The second one was designed as a copy of the first, whereas the first simply continued to exist. Yes, they are very much alike spatially, but they do not actually have the same relationship with respect to time.

And if I set one on fire, the other will not be set alight. They are not the same object.

[Lord Torath]

Okay, but let's say you put both Lego dogs down, and someone comes and hides both of them from your view, moves them around, then reveals them again to you. Can you tell which one is the original?

[Tyndmyr]

Okay, but let's say you put both Lego dogs down, and someone comes and hides both of them from your view, moves them around, then reveals them again to you. Can you tell which one is the original?

Likely not.

There are all sorts of things I might not be able to discern, particularly if someone is intent on playing a trick on me. That does not change the nature of reality. The two dogs are distinctly different, and the fact that I do not know which is a copy does not transform them into the same object.

[georgie_leech]

Okay, but let's say you put both Lego dogs down, and someone comes and hides both of them from your view, moves them around, then reveals them again to you. Can you tell which one is the original?

Not remotely; nevertheless, one of them has a continuity of existence that the other hasn't. It's not a quality I can test for in any physical way, but seems to me that we don't really have a way to confirm object permanence either. It's always possible that this really is a simulation for one person and it just loads everything when interacted with in some way Despite that, I've never met an adult that had a problem with the idea that things continue to exist when they weren't looking at it.

[Lord Joeltion]

snipped

I admit not being too familiar with this magic you call "ablation". But that's only natural, I'm no scientist after all. If I understood correctly, it's a nice way to falsify theories and previous statements that were experimented upon and assumed to be true. But to falsify assumptions and hypothesis? Dunno, it sounds impractical or inconclusive. To me, it doesn't look like the best approach to determine anything a prima facie; without prior study. There's simply many variables you may not be aware of, many variables that you simply have not even the intuition to identify what the actual results of the experiment mean.

My first objection would be that you aren't removing any quality of me, you are removing ALL OF ME altogether. The only thing you are "ablating" is myself (as a whole) from a bigger system called "society". So, yeah... with the teleporter as a method of experimentation, in fact you are observing the impact "teleporting subject Myself" has upon society; since that is the only thing you are changing. You removed me from somewhere, then observed how society reacts. But that's the only result you would get. You still haven't removed any "piece" (or rather, any quality) from me (substantial or not) and observed any kind of change upon me. You disintegrated me, then created something back; which might as well me "resurrected" or a perfect copy from me (which, ontologically CAN'T be me, for it's very definition is being Not-Me).

My second objection goes to the part you reduce my argument to: "I have a theoretical or philosophical deduction that a teleporter making a copy of me should not copy my consciousness"; which is completely false, for that is not relevant to my argument. My argument can be summed up as "A teleporter may produce perfect copies from me, but it can't assure me that my personal-experience won't be interrupted during the process". The only way to solve that, is the discovery of a (different) fantastic technology that reads minds and is able to transmit my consciousness through virtual space (digitalization, for example). Or even better, the invention I favor: Space-Folding Machines, which lack the kind of risk I fear

My third objection is this premise: "then if you observe the fact 'people go through the teleporter and do not report any disruption in consciousness' "; it's the very conundrum where we are disagreeing. You believe consciousness is observable in any meaningful way; while I am sure we haven't accomplished yet such a tremendous breakthrough. So no, I can't observe anything because I lack the methodology to do so. And no, you don't have it either.

And lastly, when you said "But specifically because you claim to experience your consciousness, you're saying that a person can actually measure their own consciousness and exhibit the result of that measurement by e.g. talking about how they feel they're conscious", you are simply saying shenanigans. "A measure of myself" is intrinsically devoid of any kind of meaning if said measure can't be observed by a third party. People, scientist, humans; have invented the concept of "measure", not as a method of understanding per se; but as a method of communication and a way to agree upon what senses communicate to our brains.

I can't have a "measure" of myself by simple acknowledgement/experience of myself; just the same way I can't communicate to you the way I see the color red. We had to decode lenghtwaves and label them as codes like this: #FF0000 to have a meaningful understanding of the science of colors. It's a fundamental problem of both language and science since "day 1". Just because I experience something, it doesn't mean I can "measure" it. "Measurement" implies communication; and self-awareness can not be communicated.
I have no way to convince you I am self-aware and not a killer robot from the future programmed to assassinate Justin Biever. Unless you are terribly gullible, of course. Are you? don't tell Justin

If we could manipulate things with subatomic precision medicine wouldn't be so fiddly either

Well, psheeaah. But if we could, it would also mean that even the chaotic chemistry within us will be relatively less chaotic. Then again, there are still many mysteries of how the body actually works, despite the level of chaos at cell levels. Many drugs we haven't yet discovered how they actually work; or why. Even Viagra wasn't intended to be used as an "ego"-enhancement drug. I think most of the restrains of medicine are purely about technology; but some of it is also certain lack of research. Medicine is kinda old, but biochemistry and "modern" methods of medicine are relatively young.

Despite that, I've never met an adult that had a problem with the idea that things continue to exist when they weren't looking at it.

This one stole me a chuckle

[NichG]

They don't, though.

Look at the two lego structures. The second one was designed as a copy of the first, whereas the first simply continued to exist. Yes, they are very much alike spatially, but they do not actually have the same relationship with respect to time.

This is why we very carefully developed the 'idem' concept. When A => B, that's more than just 'B comes from A', its 'in every way that it is possible for B to vary, that variation is caused by the same variation in A and only by that'.

If somehow matter internally varied depending on the specifics of its own temporal continuity in a way that could cause a difference in events in how it interacts with the rest of the universe, A => B means that B has a copy of that too.

That's overkill, because matter doesn't record its own history or temporal continuity or anything like that. But if it did, either the lego architect makes an error that causes the second lego object to behave differently with respect to some stimulus (it's not idem), or the lego architect is really good and can forge the temporal record as well (and it is directionally idem in the same way you are with your future self).

Of course the lego architect isn't brainwashing an outside observer, so an outside observer can determine a convention such as 'I'm going to ask the architect which one the fake is, and thats the fake' for example. But that has nothing to do with the object itself.

And if I set one on fire, the other will not be set alight. They are not the same object.

If you set your present self on fire, your past self will not be alight. If you set your past self on fire, your present self will be alight.

If you set the object exiting the teleporter on fire, the object entering it will not be alight. If you set the object entering the teleporter on fire, the object coming out of the teleporter will be alight.

I admit not being too familiar with this magic you call "ablation". But that's only natural, I'm no scientist after all. If I understood correctly, it's a nice way to falsify theories and previous statements that were experimented upon and assumed to be true. But to falsify assumptions and hypothesis? Dunno, it sounds impractical or inconclusive. To me, it doesn't look like the best approach to determine anything a prima facie; without prior study. There's simply many variables you may not be aware of, many variables that you simply have not even the intuition to identify what the actual results of the experiment mean.

It's actually really simple and really solid. If I cut off your leg and you can still talk, your leg was not a necessary condition for you being able to talk. There's no 'but you don't really know...' or 'but the variables, what about the variables?!'.

My first objection would be that you aren't removing any quality of me, you are removing ALL OF ME altogether. The only thing you are "ablating" is myself (as a whole) from a bigger system called "society". So, yeah... with the teleporter as a method of experimentation, in fact you are observing the impact "teleporting subject Myself" has upon society; since that is the only thing you are changing. You removed me from somewhere, then observed how society reacts. But that's the only result you would get. You still haven't removed any "piece" (or rather, any quality) from me (substantial or not) and observed any kind of change upon me. You disintegrated me, then created something back; which might as well me "resurrected" or a perfect copy from me (which, ontologically CAN'T be me, for it's very definition is being Not-Me).

I'm ablating your body while maintaining the pattern of organization of the particles that composed your body at one point in time. If it turns out that that pattern is actually all you are or ever were, then it turns out I didn't ablate you. If it turns out that there's more to you than that pattern, then I ablated you.

Your 'theory' here that we're testing is this definition by which "which, ontologically CAN'T be me, for it's very definition is being Not-Me". According to that theory, 'you-ness' is stored in a degree of freedom the teleporter doesn't transport. It's not the only theory you could have - a competing theory for example would be one in which the ontology is different and identity doesn't distinguish between things and perfect copies of them.

The prediction of your theory is that when you teleport, 'you-ness' must be lost. Your theory also doesn't allow you-ness to not be lost.

If the evidence on performing the experiment is that behavior is entirely unaffected, then that means exactly one of two things:

1) The theory is wrong (most likely in the axiom which assumes that ontological identity matters)

or

2) 'You-ness' has nothing to do with behavior

However, if its 2), then that raises the question - when you talk about how you're you and how that matters to you and how you'd know if it ended and so on, where is that speech coming from? That speech is part of your behavior, after all. If 'you-ness' has no effect on behavior, and yet you claim to be aware of it, then the only possibility is that you are not telling the truth (either because you yourself are deceived, or willfully).

My second objection goes to the part you reduce my argument to: "I have a theoretical or philosophical deduction that a teleporter making a copy of me should not copy my consciousness"; which is completely false, for that is not relevant to my argument. My argument can be summed up as "A teleporter may produce perfect copies from me, but it can't assure me that my personal-experience won't be interrupted during the process". The only way to solve that, is the discovery of a (different) fantastic technology that reads minds and is able to transmit my consciousness through virtual space (digitalization, for example). Or even better, the invention I favor: Space-Folding Machines, which lack the kind of risk I fear

That just buries your deduction - when you claim that its possible that a perfect copy could be made, but still wouldn't copy your personal experience, that relies on a particular view about what personal experience is. Imagine if I said 'it may produce perfect copies of this statue, but it can't assure that the leg will be in place correctly'. In order for that not to be nonsensical, I have to accept that 'the leg of a statue is not a priori part of what is implied by a perfect copy'.

Since I'm not willing to accept that particular deduction just as a matter of axiom, I'm objecting: 'a perfect copy, by definition, copies everything - including your consciousness. If it doesn't, it wasn't a perfect copy.'

It would be different if you said e.g. 'well, I don't think a teleporter that makes a perfect copy can ever be built', because while that's denying the premise we could at least say 'alright, lets talk about teleporters that can be built' or 'I take your point but I'd rather talk about what if a perfect teleporter could be built'. But you're saying 'under the premise that the teleporter is perfect, (the teleporter is not perfect)'.

My third objection is this premise: "then if you observe the fact 'people go through the teleporter and do not report any disruption in consciousness' "; it's the very conundrum where we are disagreeing. You believe consciousness is observable in any meaningful way; while I am sure we haven't accomplished yet such a tremendous breakthrough. So no, I can't observe anything because I lack the methodology to do so. And no, you don't have it either.

I hold that either:

- Consciousness is observable in a meaningful way
- Consciousness doesn't actually exist

Take your pick.

If we replace the word 'consciousness' with, say, 'vim' or 'vital force', it's the same story. If someone comes up to me and tells me that unless they eat a spoonful of tictacs every day, their vim will be depleted, then either there is something which they are detecting which actually does have this relationship with tictac consumption or they're delusional and made something up. Either 'vim' is observable, or its imagined. Consciousness doesn't get a special pass - until its observed, its just another random word.

And lastly, when you said "But specifically because you claim to experience your consciousness, you're saying that a person can actually measure their own consciousness and exhibit the result of that measurement by e.g. talking about how they feel they're conscious", you are simply saying shenanigans. "A measure of myself" is intrinsically devoid of any kind of meaning if said measure can't be observed by a third party. People, scientist, humans; have invented the concept of "measure", not as a method of understanding per se; but as a method of communication and a way to agree upon what senses communicate to our brains.

I can't have a "measure" of myself by simple acknowledgement/experience of myself; just the same way I can't communicate to you the way I see the color red. We had to decode lenghtwaves and label them as codes like this: #FF0000 to have a meaningful understanding of the science of colors. It's a fundamental problem of both language and science since "day 1". Just because I experience something, it doesn't mean I can "measure" it. "Measurement" implies communication; and self-awareness can not be communicated.
I have no way to convince you I am self-aware and not a killer robot from the future programmed to assassinate Justin Biever. Unless you are terribly gullible, of course. Are you? don't tell Justin

You can't convince me, but you definitely can talk about your self-awareness. You've been doing so. Either that's deluded, made up stuff, or its actual evidence of you detecting your own self-awareness. I'm actually assuming the second, which is the more generous option. But under the assumption of the second case, I'm not going to give you privileged status in being trusted in talking about your self-awareness - if someone else talks about theirs, I have to assume that what they say holds exactly equal weight.

If I just assume that you're deluded, this entire conversation gets much, much easier in some ways. I can then say 'it doesn't matter if you the teleporter preserves consciousness, because consciousness doesn't exist, and there's no 'you' to be killed in the first place'. But I find that while that is also a theory consistent with our observations, its much harder to have a civil conversation with people while holding it.

[Lord Torath]

Note that "detecting something" counts as measuring it. That might be the only thing you can measure about it, but that is still a measurement.

[Tyndmyr]

Not remotely; nevertheless, one of them has a continuity of existence that the other hasn't. It's not a quality I can test for in any physical way, but seems to me that we don't really have a way to confirm object permanence either. It's always possible that this really is a simulation for one person and it just loads everything when interacted with in some way Despite that, I've never met an adult that had a problem with the idea that things continue to exist when they weren't looking at it.

Strictly speaking, if we're down to "observing every molecule" levels of science, it may in fact be possible to observe the past of an object.

I mean, we do it every time we look at the stars. Arbitrarily higher resolution of observation(a necessary conceit for the teleporter)* allows us to resolve arbitrarily smaller objects at arbitrary distances.

Therefore, there'll be points at which observers could view the past history of objects, rather than their present. The idea of "you can't tell the difference between the items" falls apart. Someone at the right point in space could totally tell them apart due to observing their past.

Ultimately, the whole problem exists because people view the time dimension as somehow "not counting". But spacetime is ultimately unified, and thus, it is not even theoretically different from placing a cloth over an object and claiming it now no longer exists. This literally *is* an issue of object permanence.

*In reality, uncertainty poses a huge problem for doing this, probably rendering both such observation and the teleporter itself impossible.

[Lethologica]

It seems undisciplined to me to make definite statements about this thought experiment when it is not possible even in principle. We can't even know the exact nature of even much simpler objects, even in theory - never mind what happens when the 'even's are taken away. Given that, we can add whatever axioms we want to enable this kind of teleportation, and what outcome we get derives directly from those axioms.

For example, I can postulate that the reason we can perform this teleportation is because we've figured out how to transfer the consciousness, which contains a record of the physical form (we don't know what it is, we just know that the consciousness knows, like proteins 'know' how to fold), which can then be reconstructed. No consciousness was obliterated in the making of this film.

Also, I think wondering which is the original in the duplication case is rather moot. They could both be the original in the sense of maintaining continuity of consciousness, because who can say that's impossible when we just did the impossible by making that person in the first place, and the same issues would still arise when it comes to apportioning their material goods, social obligations, etc.

Also:

Ultimately, the whole problem exists because people view the time dimension as somehow "not counting". But spacetime is ultimately unified, and thus, it is not even theoretically different from placing a cloth over an object and claiming it now no longer exists. This literally *is* an issue of object permanence.

Then let's look at another scenario: how do we know that someone woken up from cryosleep is the same person? The time dimension isn't special, so why should this time-transferred consciousness be any more continuous than the space-transferred consciousness of the teleported person?