I'll first adress basho's arguments:
a.)
Only if you are running the simulation within the same simulation, trying to reproduce the environment with the same fidelity.
Concurred. Smaller than the actual universe's simulations are conceivable.
b.)
We already run simulations because they are far more cost efficient or safe. Car crash testing for example... its far cheaper to crash thousands of simulated cars. And nuclear weapon testing is safer in simulation. And then there are things like simulations of supernova. I'd prefer a simulation in such a case.
Here we are talking about totally different kinds of simulations. And actually with your examples you prove exactly what I'm trying to say. I was implying to a simulation which would simulate every aspect of reality. The simulations you present simulate only a minute fraction of actual reality (e.g. car safety tests). Cost-efficiency would refer to the ridiculousness of creating a simulation which contains unnecessary aspects of reality when trying to run tests within a narrow topic.
c.)
Thoughts and feelings have been empirically and practically correlated with the neuronal underpinnings. For example, during brain surgery, poking and stimulating different parts of the brain elicits very specific memories and feelings from the patient.
That is only a very small portion of correlating neuronal events with actual cognition. The events which take place inside the brain and give rise to consciousness occur so fast (millions of events per milliseconds) that asking a person what did you think time t1 and then matching the description given by that person to the neuronal events observed during time t1 in practice is virtually infeasible. First of all because we ourselves are never aware what exactly gave rise to the thoughts experienced and what EXACTLY was the milieu we were feeling at that time t1. Every passing moment of consciousness carries enormous amounts of information (even though we are not always aware of that) and after every passing moment the information becomes more and more distorted (because of memory), so that correlation feat becomes tremendously difficult (and if I may say in practice, impossible).
d.)
How many different advanced, intelligent species does your sample contain?
Ok, my bad, I should have used the word "individuals" instead of species. You got me there.
e.)
The Sims? Second Life? World of Warcraft? Star Wars Galaxies?
Games would not require creating all aspects of our observed reality, but yes that is rather a matter of question when we would debate whether our current reality is simulated or not.
Live Forever's post:
a) Perhaps, but perhaps not. Bostrom makes a plausible argument (3rd section) as well as many other places that describe the computational requirements for simulating a universe, most assuming quantum computational power, as manageable at some future point. (also note, Bostrom's isn't the only theory, there are several others). Also, who is to say that the simulation we might be living in (or might create) has to be the same size as the original? That is not stated as a requirement. This type of argument in general strikes me as the same type of argument as the people who say a brain can never analyze itself to the point of fully understanding itself. I think that we will eventually be able to simulate entire human brains to the point of individual neurons, and so that argument seems illogical to me.
Mostly agree, but then again my view is that the brain cannot analyze itself to the bone, so to say. I explained my supposition above. And because it seems illogical to you does not make it illogical in reality.
b) There are a lot of things now that are simulated (atomic bomb explosions, galaxy interactions, car durability testing, etc. etc.). They are all far cheaper to do in simulation form than in real life. As the cost of computing falls, it will only increase the amount of things that are simulatable. (some have stated that it will be a great way to find new drugs when we can simulate the entire human body and the drugs interactions on it) We find tons of things out now using simulations, with even more on the horizon. To say this will stop at some point in the future is illogical. (to me at least)
Like basho's and your simulations concern only certain aspects of reality which would be productive to simulate.
c) We have models that are good now, but in the future it should not be any problem to simulate a human body, mind, etc. You are arguing against a human level AI in this argument, and I suppose this is just a matter of opinion, but I would suppose the majority of people on this site (many of who work in AI fields attempting to create AGI) would disagree with you on this point.
Yes, I do not share the view that (exact) humanlike AI is possible. I have my reasons.
d) Again, to place our morals on another species is disingenuous. You are stating that those that run a simulation must necessarily be like the traditional "all good" view of "God". To place our moral framework on those that are necessarily much more advanced is not something we can do. To say that when we get to the point of being able to run such simulations that we won't allow ourselves to run accurate simulations just out of some sense of "moral obligation" is a bit illogical to me as well.
Again I was following a little different train of thought. Technology itself does not make anyone moral or "good", but the very fact that one would create a real-life simulation entails that very someone knows practically everything about life, evolution, consciousness etc. Why to me this generates moral and ethical paradox is because it would contradict such a being's "empathy". If you program things like death and suffering and in a way that we can observe those phenomenon in our world, would mean that the being must understand what it means to suffer (in order to realistically simulate it). This awareness leads to the very basic building blocks of empathy which is "knowing/feeling what the other finds unpleasant". Unless the creator of the simulation is a sadistical psychological mess (and I can't imagine that a being with distorted psychology would be capable of creating a realistical representation of an objective reality), then it would very logically find creating a simulation like our world, a very sick idea.
e) Why do we simulate the universe now on supercomputers? The logical motive is to see how things work, and to indulge our curiosity. To say we will at some point get tired of learning new things and expanding what we know is illogical.
In my opinion even curiosity knows its limits. Even though I'm interested what my neighbour next-door is doing in his/her house, I won't go placing videocameras into his/her house to satisfy my curiosity (although I can imagine some people could do that, but creating a simulation would be a collective effort and thus makes it improbable).
Edited by ikaros, 16 May 2007 - 11:30 AM.