51–60 of 62 posts
Earlier quoted context omitted.
Universe doesn't have to be, just neurons and other brain hardware. Of course you can always see quantum phenomena affecting outcome, you just have to rule those out as neligible / much lower than noize threshold.
There are many functions which cannot be computed on a Turing Machine. https://en.wikipedia.org/wiki/Computable_function#Uncomputab...
I find the metaphor bad. My brain might work like a computer (computation et al), but the word computer evokes actual existing computers to a layman -- and my brain doesn't have a von neumann architecture, nor does it run something similar to our software. A neural network does computation of course, but we wouldn't call it a "computer" in the sense we use the word when dealing with, well, computers.
> and my brain doesn't have a von neumann architecture "Computer" doesn't automatically equal "von neuman architecture". > nor does it run something similar to our software A lack of software-equivalent is a pretty good criticism, though.
No, but that's what most people are familiar with (if they are familiar with computer architectures in the first place), so it pretty much equals that for them.
Earlier quoted context omitted.
> Computers are designed to interpret or translate the messy analog world where nothing is exactly 1 or 0, into a precise, deterministic digital world where everything is either 1 or 0. And I think it's highly likely that the brain does not work in this manner whatsoever But neurons totally do work by encoding information with discrete signals. https://en.wikipedia.org/wiki/Action_potential#Process_in_a_... Action po…
Though, not all neuroscientists are on the same page nor agree with the claims some people want to make… "The frequency of action potentials (APs) generated by neurons is correlated with different events; however, it does not mean that such events are ‘encoded’ in the brain using the firing rate or any other ‘temporal code ’ (e.g. interspike interval)… …Simultaneous firing of action potentials (synchrony) in a neuron…
However, the relevance of Turing model is questioned even in case of present-day computing [33] [34]. Indeed, any computing machine that follows a Turing model would be highly inefficient to simulate the activity of biological neurons and experience an increased slowdown. Since the super-Turing computing power of the brain has its origins in these ‘strong’ interactions that occur inside neurons, current models have missed the most important part. Simply, Nature doesn’t care if the N-body problem has analytical solutions [36] or can be simulated in real time on a Turing machine [37].
...
While previous models have attempted to represent Hamiltonians using Turing machines [35] the paper [1] shows that the Hamiltonian model of interaction can represent itself a far more powerful model of computation. Turing made an important step forward; however, there is no need to limit natural models of computation to Turing models. In this sense, the new framework of computation using interaction is universal in nature and provides a more general description of computation than the formal Turing model. In other words God was unaware of Turing's work and has put forward a better model for physical computation in the brain.
Earlier quoted context omitted.
Though, not all neuroscientists are on the same page nor agree with the claims some people want to make… "The frequency of action potentials (APs) generated by neurons is correlated with different events; however, it does not mean that such events are ‘encoded’ in the brain using the firing rate or any other ‘temporal code ’ (e.g. interspike interval)… …Simultaneous firing of action potentials (synchrony) in a neuron…
Wow, fantastic paper: However, the relevance of Turing model is questioned even in case of present-day computing [33] [34]. Indeed, any computing machine that follows a Turing model would be highly inefficient to simulate the activity of biological neurons and experience an increased slowdown. Since the super-Turing computing power of the brain has its origins in these ‘strong’ interactions that occur inside neurons,…
The amount of money being wasted trying to make special FMRI's and custom EEG caps is lost on me when attempting to solve equations along variable boundary conditions seems like it should come first…
Here's the paper the author wrote which that paper is branched from: http://www.researchgate.net/profile/Dorian_Aur/publication/2...
Earlier quoted context omitted.
The point is that out definition of the computability of a function is too narrow, and does not take into account the real-time all connected nature of the computation the brain does. Therefore, we lack formal way of satisfactorily compare the brain to a Turing Machine.
I don't see why. Your computer reacts to keyboard and mouse in the real time, also for network packets going in and out. And it's plain turing machine. It does so by regarding every input as they become available, one by one. Information comes to brain via neurons, and those have finite speed, it's not as they had quantum-like properties. Signal in; signal out, with delay and altered magnitude.
I, too, think that the brain is 'a computer'. But the formal definitions of computation we have today require the computer to be left alone for some time between the moment the input is given and the moment the output is due. The brain does not work like that, ergo we lack a formal model of how the brain operates, and thus the definition of 'computable function' -which is based on our model of computation- is less general than what the brain can do.
e.g. We have no proof that the only way neurons communicate with one another is via the synapses.
Earlier quoted context omitted.
I don't see why. Your computer reacts to keyboard and mouse in the real time, also for network packets going in and out. And it's plain turing machine. It does so by regarding every input as they become available, one by one. Information comes to brain via neurons, and those have finite speed, it's not as they had quantum-like properties. Signal in; signal out, with delay and altered magnitude.
The important word is : formal. I, too, think that the brain is 'a computer'. But the formal definitions of computation we have today require the computer to be left alone for some time between the moment the input is given and the moment the output is due. The brain does not work like that, ergo we lack a formal model of how the brain operates, and thus the definition of 'computable function' -which is based on our…
I fail to capture the difference.
The distinction between analog and digital computers doesn’t seem to be so important to me. Any computation that an analog computer can do, can also be done with digital computers and vice versa. Digital computers can process continuous-time signals as precise as analog computers, and analog computers can do logic. Which kind of computer you use depends on convenience with respect to the resources and computational problem in hand.
The “brains don’t download software” part also isn’t convincing. Brains are capable of manipulating symbols, but the rule set needs to be learned in the case of arithmetic, or selected, in the case of natural language according to inputs. Also, software doesn’t have to be like “1. Do this, 2. Do that, 3. End”, like the article implies. The brain can be running on a game loop, planning future actions according to some stochastic utility function.
Also, I find it useful when thinking about brains to separate the functions that are done unconsciously from those that are not. To me, unconscious functions of the brain seem more like using specialised hardware. Of course, those parts are almost certainly doing computation to full their functions, but it is “not abstract symbol manipulation according to some learnable rules” that people associate with consciousness.
Earlier quoted context omitted.
> Computers are designed to interpret or translate the messy analog world where nothing is exactly 1 or 0, into a precise, deterministic digital world where everything is either 1 or 0. And I think it's highly likely that the brain does not work in this manner whatsoever But neurons totally do work by encoding information with discrete signals. https://en.wikipedia.org/wiki/Action_potential#Process_in_a_... Action po…
More directly: https://en.wikipedia.org/wiki/Action_potential#.22All-or-non...
As such it is transmitting a fourier transform of the signal it is transmitting between the cells (but with different basis functions, not cosines). When a pulse is received, this will result in a sudden jump of the action potential, and then slowly die down. If a second pulse comes while the first is still in action it gets multiplied by the current value of (this particular) input potential. Rapid pulsing (3-4 ms between pulses) means something along the lines of "multiply membrane potential by 10 for every pulse", whereas a pulse width of ~350ms means "keep the membrane potential constant". Any one of these pulses may push a neuron "over the edge" and have it transmit a pulse of it's own (or a series of pulses as pulsing the axon lowers the membrane potential by a factor. This may or may not cause it to go below the trigger level, and so sometimes the axon needs to pulse multiple times to get the membrane potential back below the trigger).
This is a beautiful way to encode simplifications of analog signals. It's log scale in the sense that 2 rapid pulses encode that the function value should be an order of magnitude larger than 2 slow pulses. It is extremely sparse (pulsing speeds rarely exceed 1hz, and only for short times) so it is quite energy efficient. Your brain quite literally does nothing at all unless there's actual information to process. I find it a bit misleading that wikipedia says that the frequency of pulses encodes the info : it's true, but makes one think this frequency is constant, when in fact it is changing rapidly. Does the word frequency really make sense when any 2 measurements of pulse frequency will be different ?
Normally one would call pulses digital, because they are 1 or 0. Values in between are interpreted as being either 1 or 0, definitely not analog. But the time spacing between the pulses is what matters, and this is measured to astonishing accuracy by a chemical reaction. A measurement like that, one would normally call analog. So it doesn't really match either signal type ... It is very different from any analog or digital signal I know about.
One might say your neurons contain an encoded, but uncompressed mp3 of the signal they're transferring between neuron membranes : it's transmitting in the frequency domain, but without the compression that reduces the zeroes, so it's sending very long series of zeroes, interspersed with the very occasional one. This works well because transmitting a zero is done by doing nothing at all.
Earlier quoted context omitted.
Wow, fantastic paper: However, the relevance of Turing model is questioned even in case of present-day computing [33] [34]. Indeed, any computing machine that follows a Turing model would be highly inefficient to simulate the activity of biological neurons and experience an increased slowdown. Since the super-Turing computing power of the brain has its origins in these ‘strong’ interactions that occur inside neurons,…
I thought it was fantastic as well (because you know a theory that tries to incorporate our understanding of matter and energy [which the brain is a subset of] in a field with a lot of hand-waving going on, shouldn't be that radical…), though for most neuroscientists it is out of reach because for the most part, have little to no understanding of electromagnetism (let's set aside the QED versions of Maxwell's equatio…
Earlier quoted context omitted.
> and my brain doesn't have a von neumann architecture "Computer" doesn't automatically equal "von neuman architecture". > nor does it run something similar to our software A lack of software-equivalent is a pretty good criticism, though.
> "Computer" doesn't automatically equal "von neuman architecture". No, but that's what most people are familiar with (if they are familiar with computer architectures in the first place), so it pretty much equals that for them.
They're not. There's really only two kinds of people here—those who are unfamiliar with computer architecture (99% of the population), and those who are familiar with Von Neumann architecture, and know that it's not the only game in town.