A couple CEOs (of an electric car company and a bank) recently announced things like "There's an 80% chance that we're living in a computer simulation." I'm not even going to quote the story correctly or read it because it's nonsense! These are non-scientists making these claims, so it's no surprise that the claims are pseudoscience at best.
The problem with this type of claim is that it's not based on any evidence, but rather on the logically unsound conclusion that if something observed has similarities with something known, then the observed thing must be the same as the known thing. Similarly, atoms can be modelled as spheres around other spheres, just like star systems and galaxies etc. Therefore there is an 80% chance that star systems are the atoms of some larger world, because they're similar. A turtle's back looks a bit similar to the Earth, therefore there's an 80% chance that we live on a giant turtle. One of the only creative things we know of is humankind, therefore it must be that everything we don't understand the origin of has been created by a being that looks like us (ie. that we were created in its image). And: measurements of the universe have some quantum properties, similar to computers (which we understand more), therefore the entire universe must be a computer. The computer simulation hypothesis is not so different from the god delusion or from the plethora of crackpot theories that link any marginally similar phenomena ("my EtherParticles theory explains gravity because gravity restricts movement away from mass, just like trying to move through a dense soup of particles in the ether is predicted by me to restrict movement," etc. etc. etc. etc.)
If the universe is so certainly a computer simulation in some extra-universal world, then what is that world? Why would that world exist "in reality" if ours is so certainly simulated? Wouldn't it mean that it's nearly certain that that world is also a simulation in another world, and so on ad absurdum? And where is the evidence of any of that? There is exactly as much evidence of a turtle that the universe sits upon, as there is of a computer running us.
We must be careful to speak of what the evidence says, and not confuse that with what we imagine it to mean. Extra-universal turtles, universe simulators, alternate realities where the laws of physics are anything we can imagine, are all flights of fantasy. If you have a fantastic idea, and want to speak of it being real, you must find a way to test it. If a test tells you that the universe is similar to a computer simulation, that doesn't mean it is one. You must show that it can't be anything other than a simulation, if you want to be certain that it is. And, "I can't imagine anything else it could be," is not nearly adequate reasoning. In science, unknowns stay as unknowns until there is testable theory to say otherwise. Ruling out everything but what we think we understand, is unscientific and outdated by a few centuries.
What test has been proposed by these CEOs, that could indicate that the universe is a simulation? How do those tests rule out that it could be anything else?
Tuesday, September 20, 2016
Wednesday, January 20, 2016
Cheating on the Turing Test
Continuing from an earlier post...
If a system merely mimics a human, but does so consistently, it may be called intelligent, because it demonstrates intelligent behaviour. You don't need to crack it open and see if it's actually really intelligent or just behaving so, just as we can't crack open a human to see if it's really intelligent or just behaving so.
Suppose that you have a machine with a human inside, and all the machine does is copy the human's behaviour. It behaves as a human, intelligently. The machine without the human is not intelligent, but the whole system is. For example, an old telephone with a human on one end can pass the Turing test, but the telephone on its own can't.
What happens if you have a machine that brainlessly copies or transmits a human's behaviour, but is first separated from the human before it demonstrates that behaviour? Such a thing might not pass a Turing test, but it might be made to behave as a human for as long as necessary, and could be made without intelligence at all, just a behaviour copier.
That would be a poor demonstration of artificial intelligence, and I think it's similar to what today's Turing test candidates are doing. The best Turing candidates that I'm aware of essentially access huge databases of existing human responses, and derive their responses from that. It would be like a machine with thousands of humans in it, brainlessly selecting from the humans' responses. Of course, to do that with AI it needs to be at least clever or sophisticated. But still, the behaviours the AI is demonstrating were copied from a human. They're human behaviours, with the human separated from the copying machine.
Therefore it would be pointless to say such a machine reliably acted as a human. It merely transmitted the actions of humans. I do not think that beating the Turing test that way has anything to do with machine intelligence.
On the other hand, whatever argument can be made against such a machine, can probably be made against a human. Humans are literally human-copying machines, and there's no way to say that it's impossible for a human to go through life without an original thought. One might be able to merely copy what has already been done. If one complains of a machine, "that's not enough to demonstrate true intelligence", the same can be said of a human.
If a system merely mimics a human, but does so consistently, it may be called intelligent, because it demonstrates intelligent behaviour. You don't need to crack it open and see if it's actually really intelligent or just behaving so, just as we can't crack open a human to see if it's really intelligent or just behaving so.
Suppose that you have a machine with a human inside, and all the machine does is copy the human's behaviour. It behaves as a human, intelligently. The machine without the human is not intelligent, but the whole system is. For example, an old telephone with a human on one end can pass the Turing test, but the telephone on its own can't.
What happens if you have a machine that brainlessly copies or transmits a human's behaviour, but is first separated from the human before it demonstrates that behaviour? Such a thing might not pass a Turing test, but it might be made to behave as a human for as long as necessary, and could be made without intelligence at all, just a behaviour copier.
That would be a poor demonstration of artificial intelligence, and I think it's similar to what today's Turing test candidates are doing. The best Turing candidates that I'm aware of essentially access huge databases of existing human responses, and derive their responses from that. It would be like a machine with thousands of humans in it, brainlessly selecting from the humans' responses. Of course, to do that with AI it needs to be at least clever or sophisticated. But still, the behaviours the AI is demonstrating were copied from a human. They're human behaviours, with the human separated from the copying machine.
Therefore it would be pointless to say such a machine reliably acted as a human. It merely transmitted the actions of humans. I do not think that beating the Turing test that way has anything to do with machine intelligence.
On the other hand, whatever argument can be made against such a machine, can probably be made against a human. Humans are literally human-copying machines, and there's no way to say that it's impossible for a human to go through life without an original thought. One might be able to merely copy what has already been done. If one complains of a machine, "that's not enough to demonstrate true intelligence", the same can be said of a human.
Sunday, January 10, 2016
Conjecture: Atoms are not entities
Edit, 2.5 months later: Sometimes I don't care if I sound like a crackpot, other times I read what I wrote and cringe. I'm like a split-personality of crackpot and anti-crackpot... the latter says that writing like in this post can be fairly useless because too much of it is vague and over-general to the point that it does not effectively communicate an idea. It merely presents an idea and then rambles around it.
tl;dr: I disagree with the statements "Matter is made up of particles; matter is made up of waves." Instead I think "Matter has properties of particles and properties of waves." Same goes for light. The distinction is 1) That doesn't mean it is waves and particles, and 2) It need not have those properties all the time, in every way meaningful.
I still like the idea but the following post is content-free.
As of today, I do not believe in the existence of atoms apart from their measured properties. Specifically, I think that matter will exhibit particular properties when measured on a quantum scale, but not otherwise.
To avoid this degenerating into a purely philosophical idea, such as "nothing exists when it is not measured to exist", which probably can't be falsified, I'll qualify the idea. I think that matter can be measured to behave not as particles in certain cases, such as in macroscopic observations (everyday human interaction with most matter) [edit: this is an example of a uselessly vague idea. The macroscopic behavior of large bits of matter is consistent with it being made of particles, and there's no point to asking "yeah but what if it's not?", and no test, at least none that I've identified], interaction with light as a wave (a glass lens bends light as though it has smooth homogeneous surfaces rather than individual particles), and the behaviour of Bose-Einstein condensates (the "particles" of the matter seem to take up the entire space of the matter, and are I think not distinguishable from each other as particles).
I think that the mainstream view of this would be that matter exists as particles, that it always is made up of particles, and that those particles exhibit different behaviours depending on how they're observed. My view is that the particles are emergent and only show up as a consequence of the measurement, and are not actually there otherwise.
I've long figured this is true for light, that it isn't made up of particles, but merely is quantified when measured. It doesn't "exist both as a wave and a particle"---its existence is best described in terms of conserved quantities, stuff that's always there no matter how you measure it, such as its energy; wavelike and particle-like nature is not conserved---it merely has measurable particular properties specific to certain measurements. For example, when measuring "where" some quantity of light energy is, it will be quantified into individual particular locations, but that doesn't make it necessary that the energy moved as those photons between places where it is measured, and certainly not that "it moves as a particle through both slits of a double-slit experiment at the same time," which is something that is not measured and is true only if the particle-like nature of light is persistent and not emergent from measurement. I believe the particle nature of light is not persistent between measurement, and I now believe the same is true of matter.
I don't know enough to make any claims, but I think that this alternative view could be made compatible with mainstream quantum mechanics, and might let other sciences more easily harmonize with quantum mechanics if they were forced to adopt it. Roughly, any 'weirdness' of quantum mechanics is not due to inherent properties of things and reality, but just quirks of how reality may be measured [edit: this is an example of over-generalizing an idea to justify a belief. The belief does not follow logically, it's just what I want the idea to mean]. If the particle nature of matter displays weird properties when measured one way vs. another, such nature and weirdness are not aspects of the matter independent of the measurements.
tl;dr: I disagree with the statements "Matter is made up of particles; matter is made up of waves." Instead I think "Matter has properties of particles and properties of waves." Same goes for light. The distinction is 1) That doesn't mean it is waves and particles, and 2) It need not have those properties all the time, in every way meaningful.
I still like the idea but the following post is content-free.
As of today, I do not believe in the existence of atoms apart from their measured properties. Specifically, I think that matter will exhibit particular properties when measured on a quantum scale, but not otherwise.
To avoid this degenerating into a purely philosophical idea, such as "nothing exists when it is not measured to exist", which probably can't be falsified, I'll qualify the idea. I think that matter can be measured to behave not as particles in certain cases, such as in macroscopic observations (everyday human interaction with most matter) [edit: this is an example of a uselessly vague idea. The macroscopic behavior of large bits of matter is consistent with it being made of particles, and there's no point to asking "yeah but what if it's not?", and no test, at least none that I've identified], interaction with light as a wave (a glass lens bends light as though it has smooth homogeneous surfaces rather than individual particles), and the behaviour of Bose-Einstein condensates (the "particles" of the matter seem to take up the entire space of the matter, and are I think not distinguishable from each other as particles).
I think that the mainstream view of this would be that matter exists as particles, that it always is made up of particles, and that those particles exhibit different behaviours depending on how they're observed. My view is that the particles are emergent and only show up as a consequence of the measurement, and are not actually there otherwise.
I've long figured this is true for light, that it isn't made up of particles, but merely is quantified when measured. It doesn't "exist both as a wave and a particle"---its existence is best described in terms of conserved quantities, stuff that's always there no matter how you measure it, such as its energy; wavelike and particle-like nature is not conserved---it merely has measurable particular properties specific to certain measurements. For example, when measuring "where" some quantity of light energy is, it will be quantified into individual particular locations, but that doesn't make it necessary that the energy moved as those photons between places where it is measured, and certainly not that "it moves as a particle through both slits of a double-slit experiment at the same time," which is something that is not measured and is true only if the particle-like nature of light is persistent and not emergent from measurement. I believe the particle nature of light is not persistent between measurement, and I now believe the same is true of matter.
I don't know enough to make any claims, but I think that this alternative view could be made compatible with mainstream quantum mechanics, and might let other sciences more easily harmonize with quantum mechanics if they were forced to adopt it. Roughly, any 'weirdness' of quantum mechanics is not due to inherent properties of things and reality, but just quirks of how reality may be measured [edit: this is an example of over-generalizing an idea to justify a belief. The belief does not follow logically, it's just what I want the idea to mean]. If the particle nature of matter displays weird properties when measured one way vs. another, such nature and weirdness are not aspects of the matter independent of the measurements.
Subscribe to:
Posts (Atom)