4 ms·

I suppose one problem is that we don't know precisely how conscious experience arises from neurons firing, which makes the question of simulation fidelity unans

I suppose one problem is that we don't know precisely how conscious experience arises from neurons firing, which makes the question of simulation fidelity unanswerable at this point. If brains are simple enough that they model N discrete binary functions from simple stimuli domains to simple response ranges implemented in a manner that a physical interaction in a region of the brain is directly mapped to a single function and small subset of relations then perhaps a BIV wouldn't have to perform physics simulations to be accurate enough, but my guess is that an "accurate enough" BIV would be slightly more complex than a physical simulation of a brain; part to do the brain stuff and another part to interpret the necessary effects of physical changes on a brain in the same state as the BIV and map them to the appropriate stimuli for the BIV. In that case the BIV is extraneous, but would exist. As an aside, there's a related thought experiment about gradually replacing the neurons in a person's brain with artificial neurons that behave identically. The question is "at what threshold is the brain a different person" and my answer is never. If the artificial neurons behave identically then the person doesn't change, but their substrate does. That would be a method to construct a BIV, but the artificial neurons would need to respond to their local biochemical environment in the same way as original neurons. It would certainly be possible to make these artificial neurons immune to physical shock or cosmic rays but this would change the internal experience of the person; they would stop being vulnerable to some forms of physical trauma, some drugs, and some physiological states. The person would notice the change of becoming a BIV, because their experience would be changing at the extremes where it matters that they are physically embedded in a dangerous environment. If the BIV was constructed "outside the simulation" with an artificial biochemical environment they would notice an even bigger shift in perception, e.g. caffeine and other drugs no longer working as expected.

mannykannot 5y ago
I see that you are thinking of the entire BIV as being a simulation, while my impression is that people are thinking of a biological brain, with "only" the electrical and chemical interface being artificial - but this is just a thought experiment, and for the purposes it is being invoked, these implementation "details" are moot. What you are thinking of seems to me to be more like Bostrom's Simulation Hypothesis[1]. Chalmers, in his new book, "Reality+', argues that there's no way to be sure we are not simulations. I don't often agree with Chalmers, but I think he is right here, and, furthermore, the simulated universe does not have to be much like the world running it, it "merely" has to be self-consistent enough that any conscious entities in it cannot deduce that they are simulated. Here's an entertaining short story on the topic: https://qntm.org/responsibility https://qntm.org/responsibility [1] https://www.simulation-argument.com/ https://www.simulation-argument.com/ Bostrom was far from the first person to hypothesize it, but it got attention from him being a prominent philosopher who based a paper on it.
- benlivengood 5y ago
  I don't think it matters if the BIV is a physical brain or a simulated brain or a spirit or soul. I also believe that an accurate simulation is indistinguishable from reality, and this is actually the crux of my argument that a BIV wouldn't work. Simple wires hooking a BIV to a simulation are not sufficient to trick it into believing it is in reality. For the BIV to not know that it's connected to a simulation, the simulation must be physically accurate. To be physically accurate, the actual brain in the vat must be simulated as if it were embodied in the simulation so that the proper and accurate feedback can be presented to it through the interface that attaches it to the simulation. Completely ignoring the physical embodiment of brains leads to dualism. Laggy, incomplete, and inaccurate simulation will cause the brain to feel disconnected from what it thought was reality. Contrast BIV with "a magic spell could make a person believe they are in the real world, but actually they are not." and see if there is any more or less explanatory power in the phrasing. Connecting a brain (how) to a simulation (of the world plus which parts of the brain?) so that it believes (how, why) that it is in the real world requires explaining the how and why and what without magic. Contrast BIV with dreaming where it is often pretty easy to tell that we're in a dream, or when we can't tell consciously that we are in a dream it still feels different when we remember it later. Dreams that are very hard to tell from reality are very accurate is my claim. Contrast BIV with the history of souls, spirits, and other supernatural explanations where over time we have determined that physical brains are what give rise to our conscious experience via a host of physical evidence and experimentation. That process of scientific discovery is precisely what would require perfectly accurate simulation of the real brain by a BIV+simulator otherwise science would reach the (correct) opposite conclusion that our BIV-experience happens outside of the physical world. EDIT: and finally, if the simulation is accurate enough then the brain in the vat is extraneous. The conscience experience will (also) be happening within the simulated brain, where the feeling of being embodied in the world will be accurate. A BIV in this case is simply a physical duplication of another brain.
  - mannykannot 5y ago
    We already have the technology to connect electronics to individual neurons and to collections of them - cochlear implants, for example. We also have the technology to keep organs alive outside of the body. We can simulate the world with enough fidelity to make CGI in movies seem realistic. None of these are advanced to the point where a BIV is feasible today, but it would take much more than I have seen so far to persuade me that they are not feasible in principle (on the other hand, if you are only arguing for the former, I agree.) When you say an accurate simulation is indistinguishable from reality, we need to be careful about what we mean by 'reality' here. The reality here is that the BIV is a simulation, so unless the BIV knows this, its experiences are not reality. > For the BIV to not know that it's connected to a simulation, the simulation must be physically accurate. Why? It would be sufficient, but why would it be necessary? The BIV does not know what reality is like, and nor does it know what reality is like for those who put it in the vat. I imagine that its neural architecture has been shaped by evolution to preconfigure it to comprehend a world like ours, but we can tell that exact fidelity is not necessary because people with various congenital or neonatal sense-limiting pathologies still make very good sense of the world (Helen Keller is the go-to example.) > If the simulation is accurate enough then the brain in the vat is extraneous. For us, that is probably true, but there are philosophers and scientists who think there is something special about brains that makes them impossible to simulate, and so presumably would see a difference here.

mannykannot 5y ago

I see that you are thinking of the entire BIV as being a simulation, while my impression is that people are thinking of a biological brain, with "only" the electrical and chemical interface being artificial - but this is just a thought experiment, and for the purposes it is being invoked, these implementation "details" are moot. What you are thinking of seems to me to be more like Bostrom's Simulation Hypothesis[1]. Chalmers, in his new book, "Reality+', argues that there's no way to be sure we are not simulations. I don't often agree with Chalmers, but I think he is right here, and, furthermore, the simulated universe does not have to be much like the world running it, it "merely" has to be self-consistent enough that any conscious entities in it cannot deduce that they are simulated. Here's an entertaining short story on the topic: https://qntm.org/responsibility https://qntm.org/responsibility [1] https://www.simulation-argument.com/ https://www.simulation-argument.com/ Bostrom was far from the first person to hypothesize it, but it got attention from him being a prominent philosopher who based a paper on it.

benlivengood 5y ago
I don't think it matters if the BIV is a physical brain or a simulated brain or a spirit or soul. I also believe that an accurate simulation is indistinguishable from reality, and this is actually the crux of my argument that a BIV wouldn't work. Simple wires hooking a BIV to a simulation are not sufficient to trick it into believing it is in reality. For the BIV to not know that it's connected to a simulation, the simulation must be physically accurate. To be physically accurate, the actual brain in the vat must be simulated as if it were embodied in the simulation so that the proper and accurate feedback can be presented to it through the interface that attaches it to the simulation. Completely ignoring the physical embodiment of brains leads to dualism. Laggy, incomplete, and inaccurate simulation will cause the brain to feel disconnected from what it thought was reality. Contrast BIV with "a magic spell could make a person believe they are in the real world, but actually they are not." and see if there is any more or less explanatory power in the phrasing. Connecting a brain (how) to a simulation (of the world plus which parts of the brain?) so that it believes (how, why) that it is in the real world requires explaining the how and why and what without magic. Contrast BIV with dreaming where it is often pretty easy to tell that we're in a dream, or when we can't tell consciously that we are in a dream it still feels different when we remember it later. Dreams that are very hard to tell from reality are very accurate is my claim. Contrast BIV with the history of souls, spirits, and other supernatural explanations where over time we have determined that physical brains are what give rise to our conscious experience via a host of physical evidence and experimentation. That process of scientific discovery is precisely what would require perfectly accurate simulation of the real brain by a BIV+simulator otherwise science would reach the (correct) opposite conclusion that our BIV-experience happens outside of the physical world. EDIT: and finally, if the simulation is accurate enough then the brain in the vat is extraneous. The conscience experience will (also) be happening within the simulated brain, where the feeling of being embodied in the world will be accurate. A BIV in this case is simply a physical duplication of another brain.
- mannykannot 5y ago
  We already have the technology to connect electronics to individual neurons and to collections of them - cochlear implants, for example. We also have the technology to keep organs alive outside of the body. We can simulate the world with enough fidelity to make CGI in movies seem realistic. None of these are advanced to the point where a BIV is feasible today, but it would take much more than I have seen so far to persuade me that they are not feasible in principle (on the other hand, if you are only arguing for the former, I agree.) When you say an accurate simulation is indistinguishable from reality, we need to be careful about what we mean by 'reality' here. The reality here is that the BIV is a simulation, so unless the BIV knows this, its experiences are not reality. > For the BIV to not know that it's connected to a simulation, the simulation must be physically accurate. Why? It would be sufficient, but why would it be necessary? The BIV does not know what reality is like, and nor does it know what reality is like for those who put it in the vat. I imagine that its neural architecture has been shaped by evolution to preconfigure it to comprehend a world like ours, but we can tell that exact fidelity is not necessary because people with various congenital or neonatal sense-limiting pathologies still make very good sense of the world (Helen Keller is the go-to example.) > If the simulation is accurate enough then the brain in the vat is extraneous. For us, that is probably true, but there are philosophers and scientists who think there is something special about brains that makes them impossible to simulate, and so presumably would see a difference here.