The Simulated World According to David Chalmers
nautil.us
nautil.us
Could we not just redefine our capital R reality as a damn fine simulation? Isn’t that the message behind almost all spirituality and religion?
This pseudo-question belongs in the trash heap of philosophy along with mind-brain dualism.
It is a question generated by murky thinking—-credit to Wittgenstein for pointing out the obvious. Read Richard Rorty—any thing really but “Truth and Progress” a good place to start.
If you want way more insight than Chalmers will ever provide then read either “The Quantum Thief” by Hannu Rajaniemi, or hardcore neuroscience: “Principles of Neural Design” by Sterling and Laughlin.
This is a key point. The simulated universe is unfalsifiable, meaning that there is no experiment we can (currently, perhaps forever) perform that could prove it true or false. Even if we could conceive of such an experiment, simulation advocates could argue that the simulation was set up to produce the results we see.
Also, it's worth bearing in mind that simulation is really hard and computationally expensive: we cannot yet model a single human neuron at the atomic level, let alone at the quark level. You can try to get around this detail by saying that the simulation focusses computation / resolution on where we are looking, but this again seems to be moving the goalposts.
I would like a simulation proponent to propose a credible laws of physics for the host universe, as well as making some comments about how computation (i.e. simulation) works in that universe. For example, is the host computation subject to the same theoretical constraints that our computers are, e.g. the halting problem, and P versus NP.
[Edit] minor rewording for clarity.
> it's worth bearing in mind that simulation is really hard and computationally expensive
This is like ants saying "building mounds is resource intensive. We need our entire civilization working together to build one. You can try to get around this detail but..."
This carries the implicit assumption that if there is indeed a simulation then it is limited to human capability, which would almost certainly be wrong.
The other problem is that if this were a simulation, quantum mechanics might not be a fundamental mechanism of reality but rather arbitrary rules executed by the simulation.
Yes, totally. But the danger then is that if we say "it is more complicated than we can understand", then the simulation hypothesis essentially becomes a faith-based argument, where 'simulation' essentially means 'god', and then all the bets are off.
The scientific view is that there is a pre-existing world out there, and a Way Things Work, and if we observe and struggle enough we can understand it.
The religous/simulation approach gives primacy to the soul/consciousness, and has the world as a driver of plot.
It's top-down vs bottom-up. Do we need to simulate a trillion trillion neurons to model a single brain, or is a neuron a concept that gives scientists something to talk about?
Merely 100 billion neurons; but the real computational compartment is the synapse and its dendritic pond. That gets you up to 10 trillion miniature analog circuits operating “concurrently” in a human brain.
But all almost miraculously running without a base oscillator for para-synchrony of compute. Figure that one out and we can then start talking about general AI and simulating Neo all over again again.
Would you elaborate on this a little for those of us who don't know what a base oscillator is or what it does for a para-synchronous compute?
Analog circuits also need a time frame (phase) for computation although it is not discretized.
So where in the brain—a fancy mixed bag of salty cellular neurons and glia—is there anything analogous to an oscillator or clock?
Answer: there isn’t!
Para-synchrony is learned gradually by living and moving in a world. It is the root of the emergence of coordination of movements and eventually, of semantic sequence.
The first years of life are about building “clock systems” into neural systems by tweaking synaptic weight and connection latencies to achieve coordinated movement (a hard compute) and accurate/useful perceptions (another hard compute). Movements, sensations, perceptions are all coupled closely and share DYNAMIC modulated meta-clock systems.
We eventually control these meta-clock systems to some modest degree—and that modest degree allows us to steer attention recursively. Hofstadter makes this point in Gödel, Escher, Bach toward the end.
Neuroscientists and developmental psychologists, and even philosophers (including Heidegger) take Time with a capital T, as a given. They are fixated on Representation.
Even Piaget takes “Time” as a given. But time or synchrony as we experience it is very much a neural construct that takes years to build and fine-tune.
Piaget got very close but in the end the question and the problem eluded him like a ghost. In “The Child and Reality”, he writes on page 15 (Rosin translation): “the action of looking, of following with the eyes, the action of bringing an object to the mouth, etc.—but they all lack coordination with each other.” Yet he forgets to ask why they lack coordination.
The answer is that infants have very low and slow parasynchrony across their sensory-motor systems.
More on this topic @robwilliamsiii
I read a paper a while back that suggested the brain cortex is modular, and each module is constantly+independently comparing sensory input to expectations (arxiv.org/pdf/1810.09084.pdf).
From personal experience, when there's a lot of stimuli competing for my attention, most of it gets ignored or handled subconsciously. The flow of consciousness has fuzzy boundaries.
With computation as we know it. Is our computation universal? Also, nature "computes" in some way, whether it's simulated or not. Somehow it "knows" how to manage everything, across billions of light years.
Hence my next point , where I'm looking for more detail about how the host universe actually supports and runs the simulation.
Just consider (hypothetically, if you wish) one of our own simulations. Were there conscious, reasoning agents within it, who developed access to computation, the resources that they would have to create their own simulations would necessarily be limited by the scale and scope of the simulation we created. It would be almost inconceivable to most of them that it could be possible to "simulate" their entire world, given their understanding of physics and computation within the bounds we set up.
Then there's the matter of time. It doesn't matter very much in a discrete universe whether or not time runs "fast" or "slow" ... each discrete state is the same either way (because there is no "in between"). Just as right now it would task vast computational resources for us to simulate anything remotely like the universe we seem to live in, and it would likely not run "real time", nothing inside the simulation would be aware of the computational costs and duration in our world.
I am horrified, however, to think of people spending many hours at a time in VR or augmented reality. Nope.
Off topic, but even though I have only had one good talk with Chalmers in person, I feel some personal connection because he got his PhD with the same advisor as my friend Pei Wang, and he clashed professionally with my childhood friend Tony Brueckner (who, until he died, was chairman of the philosophy department at UCSB). I find that my enjoyment of reading books increases if I have met and talked with the author. This argues against the meta verse :-)
I am horrified that we're taking that Cartesian insight to mean we should actually put our consciousnesses inside little boxes that simulate the rest of the grand mystery of reality—because what's the difference if the lie is good enough?
It makes this atheist want to go out into nature and find the Goddess.
Accept life as a beautiful mess to fill with kindness when possible. That is pretty much Rorty’s pragmatic message.
Hack, given how weird higher dimensions are (spiky spheres etc) I wouldn't be surprised if the simulation became exponentially simpler in higher dimensions to a point where higher dimensional IT uni graduates casually simulates our universe in their equivalent of electronic pregnancy test (Show HN: I simulated toy 3d universe, it has some limitations - I had to limit the information exchange speed to 300kkm/s, it has fixed size of 93 Gly and smallest feature resolution is 1.616255e−35m).
On the other hand, you could also truly simulate each particle and molecule and then let higher-order phenomena emerge from that. For example, the universe might be an immense cellular automata.
But you could also simulate with varying precision depending on POV, which would eliminate many orders of boggle.
On the other hand, Chalmers seems to allow for consciousness of varying degrees, in the additional quality or factor that is consciousness, so maybe the whole universe is watching itself a lot of the time. (Based on the interview with him on Sean Carroll’s podcast, not reading this latest book.)
https://www.simulation-argument.com/ https://www.simulation-argument.com/simulation.html
the simulation is the entire body of scientific understanding.
the simulation is how we can do something in order to get an expected outcome, we've simulated our surroundings in order to know this.