The State of Whole Brain Emulation in 2015
creativepark.net
creativepark.net
Imagine your brain were scanned, and you found yourself running as a computer program. Consider the powers that would give you. You could make copies of yourself, obviously. You could tune some parameters, to try to increase your intelligence. You could A/B test your copies with identical memories in identical situations, to see what effect that tuning has, and you could incorporate enhancements that wouldn't work biologically. So there's a pretty high chance that you'd end up with a superhuman level intelligence, or at least John von Neumann level intelligence. Plus a large number of copies.
At the same time, the emulation might be imperfect, the modifications and enhancements might have unanticipated side effects, there's a whole lot of opportunities for sanity damage. That's on top of the base rate of humans being evil or turning evil when given power, which is itself uncomfortably high. And finally, there's one more wrinkle. Suppose brain-scanning and whole-brain emulation technology are in the wild, and a bunch of different people scan their brains and start running on computers. In this world, if you hack into someone's hardware and halt their brain program, you can repurpose it to run copies of yourself. While computer security is probably better in that future world than it is today, I seriously doubt that it will be better by nearly enough.
If you're thinking of getting into brain-emulation research, please consider also putting some work into safety research. And in the unlikely event your work should pass the level of treating mundane ailments, and moves into the territory of breakthroughs, please remember that you are playing with fire.
Are we sure we can really support that statement?
I assume you mean "isn't it possible that the human brain would run out of memory?", but I think that it depends on what you mean by Turing complete. If I hand you the specification of a universal Turing machine, then you can certainly execute it, by hand, on a limitless supply of paper—making the paper your external memory. I think that this is a reasonable way of claiming that your brain is Turing complete. (Or perhaps you worry that one must include the rest of the body to achieve Turing completeness?)
(Quick idea: recheck all calculations done so far at all computation steps that are powers of 2. So recheck everything after the first step, the second step, the fourth step, the eighth step, the sixteenth step, etc.)
If you accept physicalism, then thoughts arise as the result of electrical, mechanical, and chemical processes. As far as we know, those processes are governed by laws of physics which are all computable to arbitrary precision. Assuming that our thought processes are not infinitely sensitive to random perturbations, then the human brain can be simulated by a Turing machine.
Thus, everything we know about nature indicates that the human brain is no more powerful than a Turing machine.
Look at, for example, electron shot noise for a counterexample. Given the state of a low-output lamp and a detector, one cannot compute the state of the detector after a period of time to arbitrary precision. We can say that the detector will have x% chance of triggering within the time period - but that is all. We cannot say definitely "the detector will trigger" or "the detector will not trigger", no matter how exactly we know the initial state of the system.
And we already know that the human brain can be affected by similar processes (If nothing else, our eyes start registering at full night vision when <10 photons hit within 100ms, which is within the realm of shot noise.)
I would attack this by claiming that we have designed turing complete systems using no external input. That is to say - we can _produce_ a turing complete system, which implies we are a turing complete system given an IO port (limbs, for example).
Fundamentally I could sit and perform calculations with pen and paper.
I'm ignoring the unlimited criterion here, I assume colloquial usage.
The only real way I could fathom otherwise is by deciding to isolate the brain and consider it as a single entity. But in that case a CPU isn't really turing-complete because it can't do anything on its' own.