It turns out that the Future of Humanity Institute has a blog, and I’m being attacked there by Robin Hanson for my criticism here of the “Simulation Argument” as not science and not belonging in the NYT Science Times. In the NYT article Hanson discusses what survival strategies we should pursue in order to try and convince the Overlords of our simulation to keep us around.
In the comments here and on other blogs, Hanson and his supporters have been criticizing me for refusing to spend time answering their arguments. I just want to make clear that the reason I am not doing this is that it is possible that the Overlords read this blog, and I don’t want them to get the impression that I am willing to waste their time or mine on this kind of stupidity.
Should be: “The dismissal has nothing to do with the argument rather the subject matter itself.”
Pingback: Einstein’s contribution to the acceptance of soft/pseudo-science « Bob Dudesky
Richard wrote:
I think people are perfectly entitled to both express an opinion and choose their own particular level of engagement with arguments, or even whole classes of arguments. If, for example, every physicist had to give a detailed rebuttal to every claim for a perpetual motion machine, they would spend all their time doing nothing else, but that should not prevent them from expressing an opinion that it’s very unlikely that the claim in question is interesting, fruitful, or correct.
That said, if you really want to hear some objections to Bostrom’s argument:
(1) He invites us, inappropriately, to quantify the probability that we are living in a simulation. He is free to make conjectures and advance lines of argument that he believes make the hypothesis more likely, but this quantification is unjustified. Why? Because it is in the nature of the hypothesis that our own experience and knowledge of our simulated universe, S, are potentially completely irrelevant to the universe U in which the proposed simulation is taking place. Bostrom catalogues versions of the hypothesis where there are reasons for S to resemble U to varying degrees, but the existence of the logical possibility that no reasoning about S can tell us anything about U is enough to undermine any scheme to allocate meaningful probabilities to any version of the hypothesis.
If we decide instead to quantify the probability that our own descendants will run various classes of simulation, that’s a different matter, but rather than being intrinsically meaningless the numbers are then merely so subjective as to be useless. Like the Drake equation, maybe it’s worth writing down some of the relevant quantities initially, but once you come clean about the margins for error in some of them, you need to admit that the information content you’ve ended up with is essentially no different than saying “well, it’s not impossible”.
(2) Bostrom suggests that we must accept that either [a] civilisations will be rare and short-lived, [b] there is some reason why simulations will never take place, or [c] we are highly likely to be a simulation. He is enthusiastic in presenting arguments that seem to raise the likelihood of simulations taking place, but he doesn’t pay sufficient attention to the distinction between simulations in general taking place, and simulations of the universe as we have experienced it taking place.
There are good reasons to believe that any advanced technological civilisation will run simulations of some kind, and it’s a fact that we already simulate many interesting processes ourself: weather, population growth, the spread of pathogens, the formation of structure in the universe, etc. And if we accept as unsettled either way the question of whether computers can support consciousness, we can give that hypothesis the benefit of the doubt and allow for the logical possibility that simulations will include conscious beings. (This is, however, a step that is enough to destroy any meaningful allocation of probabilities. All we have on this issue are opinions.)
OK, so it’s logically possible that simulations will include conscious beings, but let’s look more closely at the detailed character of our own experience.
(i) Our history encompasses an enormous amount of suffering, both by humans and animals. Some of this is due to human actions, some due to “natural causes”. Intelligent beings — and most of all, intelligent beings whose ancestors we closely resemble — are highly likely to have anticipated this. Our own technological cultures are moving towards an attitude where animal experimentation, factory farming, etc. are becoming less acceptable, and (if it was accepted that beings in a simulation were conscious), the idea of a simulation that included Auschwitz, Darfur, or a few hundred thousand years of brutal evolutionary history would be utterly repugnant to most people. [Of course attempts to argue anything from our own experience are ultimately futile, but if Bostrom wants to do so in one direction, he ought really to do it in both.]
(ii) We appear to occupy a universe with a very detailed physical structure at all scales, and hence very large computing requirements. Someone interested solely in, say, our evolution or social history could easily have contrived a different cosmology and fundamental physics in order to lessen the computational burden, and though it’s conceivable that we are being selectively fed high-resolution information that does not exist when it’s not required, maintaining consistency while “cheating” would itself be computationally very difficult. And although we face the hurdle of having no knowledge of what counts as “an impractically large computer” outside the simulation, we can still note the unambiguous inefficiency of the process. Sim City does not employ lattice QCD, and however large our computers get, I suspect that it never will.
(iii) Even our own civilisation, with its limited technological and intellectual achievements, is perfectly capable of reaching useful conclusions about evolution, history, psychology, cosmology, etc., with software and other means that fall far short of simulating conscious beings. Rather than needing or wanting to simulate a complete universe, it’s likely that a more advanced civilisation would become more skilled at making intelligent computations that answer specific questions of interest efficiently. We’re asked to contemplate a civilisation with the kind of technology and resources such that uploading their own minds would be commonplace, and (certainly by the time they were capable of simulating even the Earth, let alone our whole universe) they would be likely to have a very detailed theoretical understanding of the computational basis of conscious experience. This would equip them with a host of strategies for avoiding any need to simulate us to the level where we experienced anything, whatever psychological questions they happened to be interested in. In combination with the issues of morality and efficiency, this leaves very little motivation for something like our universe to be created as a simulation.
Pro-Bostrom commentators will have a host of “Yes, but maybe …” reasons why it’s possible that my objections don’t apply. To which my reply is, “Yes, but so what?” I happily concede that it is not completely logically impossible that we are living in a simulation … but everyone knew that all along. Bostrom has contributed nothing but a few unoriginal and highly selective arguments, along with some false dichotomies and a strategy for deluding ourselves that we can allocate probabilities to these scenarios.
So, a few dim reporters have mistaken Bostrom’s waffling for a startling breakthrough in Bayesian anthropic reasoning, but if Peter Woit or anyone else thinks this is all worth no more than a despairing groan — and a complaint that it is part of a deplorable general trend — that’s entirely reasonable. I’ve spelled out some of my own reasons why I think a groan is the appropriate response, but there’s nothing intellectually wrong with just emitting the groan. Life is too short for anyone to be compelled to refute every scrap of useless nonsense at length, but that should not preclude them from calling it nonsense. Of course that judgement will sometimes be spectacularly wrong — we can all pick historical examples — but we’re each still entitled to prioritise our expenditure of intellectual energy.
This must be what is traditionally called an “epic thread”.
He [Bostrom] invites us, inappropriately, to quantify the probability that we are living in a simulation.
“You must wager; it is not optional… Let us weigh the gain and the loss in wagering that God exists… If you gain, you gain all; if you lose, you lose nothing. Wager, then, without hesitation, that He exists.”
Blaise Pascal
Did Pascal notice that there were religions in the world other than Christianity, some with gods that were much more irritable than the Christian god if you didn’t do a great deal more than merely believe in them? Maybe it just wasn’t wise to mention that quandary in his particular religious milieu. Perhaps the Future of Humanity Institute can come to the rescue and tell us the odds for being punished or rewarded for Aztec-style human sacrifice; I’m sure someone there has a rigorous Bayesian argument about this.
Hey, what am I doing giving that idea away for free … if I can interest Columbia University Press in my book Blood Sacrifice or Holy Communion: Can the Hidden Dimensions of Physics Help Us Decide? I might be up for the Templeton Prize!
From the classical paradigm
“God the Ultimate Clockmaker”
(Newtonian Physics)
to the contemporary paradigm
“God the Ultimate Video-Game Player”
what will be the next?
(There will be one, I’m just sorry I won’t live to see what it is)
Pingback: Overheard In Providence » Blog Archive » Welcome to the Matrix
At least Pascal showed some manifest interest in different cultures and their relevance towards his own discourse. Unfortunately I didn’t find good links for further examination.
About book titles:
Blood Sacrifice or Holy Communion: Can the Hidden Dimensions of Physics Help Us Decide?
Go for it. I’d also like:
Lords of other worlds. What hidden dimensions can unveil about them.
I fing Greg Egan’s comments very interesting, as I recall reading various stories by him regarding this very concept. Years ago.
Well-thought out and very good stories, too, that made me think.
This is why we need more experimentalists. Left unto themselves, theorists will contemplate their navels so deeply that they can never find their way out. Please hurry with new data before our best and brightest are lost forever.