Have you ever read one of your old blog posts or journal entries that you wrote when you were a kid, and cringed in embarrassment and horror at how stupid your post was? I have, and from what I know, it’s quite a common phenomenon, especially amongst those that read and write a lot.
I remember reading my old blog written when I was 14, and it felt particularly embarrassing. It felt incredibly pretentious and needlessly philosophical, a bunch of incredibly obvious epistemological mistakes were made; and as a result, I asked myself; wouldn’t I regret writing this blog when I’m 30 or 40 years old as well? I probably would. Hell, I feel slightly embarrassed reading most of the stuff I posted merely a year ago, nevermind a decade ago. Two years ago, I genuinely believed that I was good at writing, only realizing recently that I have a lot to improve on.
But I’ve decided to keep on writing anyway, even if I’m making a lot of foolish mistakes. Even if the probability of me regretting this is high.
It’s essentially the same problem as the reputability of science. Most of the scientific theories in the past about how the world worked was wrong, and the most intelligent people in the past were wrong about pretty much everything. They were wrong about reality, wrong about science, wrong about quantum mechanics, chemistry, philosophy, physics, wrong about politics, about the universe itself. That’s why some argue that because science has been wrong pretty often in the past, current scientific theories are probably wrong as well, even if we can’t explicitly reason why they are so.
How can science be trusted after it’s been proven wrong so many times?
How can I be trusted after I’ve obviously been wrong so many times?
The answer is that hard Bayesian evidence is a lot stronger than weak Frequentist evidence. So most of my theories about the world, and the world’s theories about the universe were wrong. Therefore using Frequentist inference, it’s true that current theories are likely to be wrong as well. But reasoning can provide a much better answer than mere correlational evidence can.
Consider for example, a turkey on a farm. For the first thousand days of its life, the farmer has given it food, and treated it exceptionally kindly. The turkey has been warned that it was in danger, for the farmer has expressed several times that for next thanksgiving, turkey will be eaten. But the Turkey disagrees, claiming that it has extrapolated those one thousand days of kindness, and therefore it has come to the conclusion that it will never be slaughtered. The Turkey is then slaughtered next thanksgiving.
Science is the same. Although correlational evidence is still evidence, it is considered a “weak” form of evidence as compared to logical reasoning from experimental data. So long as we cannot use reasoning to explain how exactly it is wrong, the probability of it being correct is still much higher than wrong.
We should trust in our current theories, because to trust in anything else would be to privilege the hypothesis, a logical fallacy. Right now, our theories are the one best guess amongst all possible ones, the explanation with the highest possibility. To discard the highest possibility answer for no answer at all is akin to saying a 70% chance of achieving something is not good enough, therefore we should switch to a method with a 0% chance of achieving something.
I should trust in my current beliefs, because if I discard them in fear of being wrong, then I will definitely not get it right. Only when I have been presented evidence that my beliefs are wrong, and evaluated them rationally to find them to be true, should I discard my beliefs, and replace them with a belief with a higher likelihood ratio.
Finally, we should trust in science because it is progressive, rather than random. Science is not so much flipping a coin to obtain a theory as it is using old theories to climb to better ones. Even wrong theories have uses — they are the stepping stones from which newer and more correct theories emerges. Just as quantum mechanics have been inspired from Newtonian mechanics, germ theory inspired from miasma, and chemistry from alchemy, it cannot be said that Newtonian mechanics and miasma should never have been invented, for that would imply that Quantum Mechanics and germ theory could never exist in the first place.
I should therefore also trust my current theories, because personal philosophical growth is not a random process. In order to obtain a worldview that better reflects the universe, it is necessary to stick my ideas on the chopping board. If that means being humiliated and having my weak epistemology exposed, all the better.
My belief in Modal Realism explained.
Over the course of my life, I have expressed my belief in Modal Realism a number of times. If you’ve never heard of Modal Realism, it essentially means that every possible universe exists, and is equally as real as ours.
Of course, to some people, this may seem like an incredibly strange belief. Some may even express fear; after all, if every possible universe exists, there is a universe where you are being tortured for all eternity. Unfortunately, reality doesn’t care what you think. Reality will continue existing whether or not you believe in something. Therefore, as the Rationality mantra goes; “If the iron is hot, I desire to believe it is hot, and if it is cool, I desire to believe it is cool. Let me not become attached to beliefs I do not want.”
I write this as an essay that I may refer others to, instead of explaining why I believe in Modal Realism every single time I mention it, which, oddly enough, comes up quite often. In the long run it will probably save a good deal of time.
Please note that the following ideas are not original. This particular essay is original, but Modal Realism has been debated for centuries, and even Turing Completeness and this “Dust theory” have been written about by people far more competent than I am, for far longer than I have existed. I do not wish to claim credit for something I have not done, and this is particularly true for this essay.
Anyway, I will attempt to “prove” Modal Realism using two axioms, which are the starting propositions that you accept in order to prove something. It is by no means a strong proof, in that it is not as strong as what most Mathematical proofs would require, but sufficient enough to assign a high probability to my ending theorem, and good enough for conversational persuasion.
My two axioms are as follow;
Firstly, that the Universe is turing-complete.
Second, that it is possible to use logic in order to create theorems from axioms.
If the universe is turing complete, you must also believe that it is possible within the laws of physics to simulate the entire universe, given infinite computational power. Nevermind the fact that infinite computational power is impossible within the known laws of physics, we shall ignore that for the moment.
This proposition makes sense because any turing complete language is capable of simulating all turing complete languages, including itself. Therefore if the Universe is turing complete, and I strongly believe it is, as we have yet to see evidence that the Universe is not turing complete, then it can also be simulated on all turing complete languages. Therefore given infinite computational power, we could theoretically simulate the entire universe as well, so long as we have the appropriate software.
If it is possible for the Universe to be simulated, you must also accept that within the simulated universe, it is not empirically possible to know the speed of which this universe is being simulated. For instance, if we were to run this universe at 0.5x speed, or 0.1x speed, it would not be possible within that universe to tell a difference.
If it is possible for the speed to vary without exerting a noticeable effect within universe, it must also follow that if the simulation of the universe is paused by a minute in the real world, before being resumed, no empirical change would result within the simulated universe as well. The same would also apply if this period were extended to an hour, a day, or even a million years.
If you accept that the simulated universe would experience no difference even if it were paused, you must also accept that it would experience no difference if the supposed “future” within the universe was simulated before the “present” within that universe was simulated. For the moment, ignore the fact that it is impossible to know the ending configuration of a simulation within first simulating the starting configuration all the way. Within your simulated universe, since each state flows smoothly to the next, it is empirically impossible to tell the difference which state is being simulated first, and which is being simulated second, therefore there is no difference.
If you accept the above argument, you must also accept that it would apply if there were three states being scrambled, instead of two. Similarly, the same would hold if there were 4, 5, a hundred, or a billion states that are randomly scrambled without regard for chronological order, and simulated accordingly.
If it is possible within the known laws of physics to simulate the entire universe, given infinite computational power, you must also believe that it is not necessary for the simulation to be run on transistors or electricity. For example, it would also be possible to run the universe on vacuum tubes, once again, ignoring the problem of computational power. Similarly, it is also possible to run the simulation on other kinds of patterns generators, so long as they can be simplified into signals consisting of zeroes and ones.
If you accept the above argument, you must also agree that the microcosmic dust and atoms of the universe can form patterns of zeroes and ones, through sheer pure randomness. For example, electrons are probabilistic fundamental particles that are dependent upon a wavefunction. We could say an electron appearing in a certain location is counted as a one, and an electron appearing in another location is counted as a zero. Therefore out of pure randomness, we could obtain a code measuring 01000100111, for example. The same logic could be applied to every particle existing througout the universe, so long as they exhibit patterns that can be classified into zeroes and ones.
If you accept that the simulation of the universe is only dependent on signals of zeroes and ones, rather than what hardware it is being simulated on, you should also accept that if the hardware running the simulation was to be cut in half, but somehow was still capable of creating the same patterns of zeroes and ones through “spooky action at a distance” (imagine a mysterious portal connecting the two parts together if you have to) , there would be no empirical difference within the simulated universe. If you accept that argument, then you must also agree with it even if the two parts of the infinite-power computer were light years apart, or if the computer was cut into 3 parts, 4 parts, or a billion parts, so long as the same pattern of zeroes and ones are being generated.
Finally, you must also agree that a simulated universe is also capable of generating bits (patterns of zeroes and ones) on its own. Since we already accept the fact that locality does not matter for computing simulations, you must also accept that a third simulation is physicically capable of being computed from a combination of a computer inside the simulation, and a computer outside the simulation.
Using the above theorems, we can say that every possible universe exists. Because every simulated universe generates its own set of bits, we can say that there are an infinite number of one bits and an infinite number of zero bits. And because the coordinates of each bit does not matter, whether it is through space or time, so long as it generates the same zero and one values in an orderly pattern, we can also say that we can rearrange those infinite number of zero and ones to form every conceivable possible universe.
Therefore: Every mathematically possible turing-complete universe exists.