Using artificial means to imitate a person's brain through neural networks?

[Unity+]

There has been some research into using MRI scans of the brain to communicate through the brain using patterns in brain scans and such(neuron interactions, etc.). Here is an idea I had.

 

If there was a way to determine how neurons interact in the brain and developing an algorithm that encompasses the interactions of neurons in the brain, and an MRI could scan the patterns that exist in the brain, could there be a way to imitate the person's mind on a computer simulation? And to be able to communicate or interact with the simulation we would have to determine how neurons interact when someone hears, speaks, tastes, smells.etc.

 

Of course, this is easier said and done, but is the basic concept sound?

[Unity+]

Does anyone have any opinion or research they want to link? I am very interested in the concept and want to learn about it.

[Fuzzwood]

So you are basically looking at this scenario? http://en.wikipedia.org/wiki/The_Schizoid_Man_%28Star_Trek:_The_Next_Generation%29

[Unity+]

So you are basically looking at this scenario? http://en.wikipedia.org/wiki/The_Schizoid_Man_%28Star_Trek:_The_Next_Generation%29

If I understand it correctly, it is somewhat like what is stated in the plot. It is merely scanning the static state of the neurons and applying to a computer simulation that contains the algorithm of how neurons interact with each other(somewhat like Conway's game of life) and runs the simulation.

[Unity+]

Isn't anyone going to respond on the matter?

[Unity+]

I have let this thread run, and I was wondering why no one is responding.

[Ophiolite]

No responses? Possibilities.

 

1) This is so speculative it is barely out of the realms of SF.

2) Genuine question: does your prior posting history - of which I am unaware - suggest that you intend to discuss a different, but related topic and no one is willing to bite?

3) It is simply answered by a single word - yes - so that discussion seems pointless.

[Unity+]

No responses? Possibilities.

 

1) This is so speculative it is barely out of the realms of SF.

2) Genuine question: does your prior posting history - of which I am unaware - suggest that you intend to discuss a different, but related topic and no one is willing to bite?

3) It is simply answered by a single word - yes - so that discussion seems pointless.

1. I am merely asking a question about a concept. I am asking if it lies in the range of science rather than speculation.
2. I am discussing the topic that I posted here.
3. I am trying to get a yes or no answer, which I haven't received.

[iNow]

We can probably simulate neural activity via a sensor (like the aforementioned MRI) and incredibly powerful computer, sure. However, consciousness itself is a bit of an emergent property of all of that activity and structure and it's not clear that it would equally emerge if run or simulated from silicon instead of biology.

 

The core challenge in your question as I see it is the concept of mind, as opposed to focusing more tangibly on concepts like structure or activity. The latter can be modeled and even approximated via simulation. The former, not so much. It's still too vague and ill-defined of a concept, IMO.

[Unity+]

We can probably simulate neural activity via a sensor (like the aforementioned MRI) and incredibly powerful computer, sure. However, consciousness itself is a bit of an emergent property of all of that activity and structure and it's not clear that it would equally emerge if run or simulated from silicon instead of biology.The core challenge in your question as I see it is the concept of mind, as opposed to focusing more tangibly on concepts like structure or activity. The latter can be modeled and even approximated via simulation. The former, not so much. It's still too vague and ill-defined of a concept, IMO.

Well, a reference point that I can give is Conway's game of life. We simply apply 4 simple rules and use two options for each square, on or off. This results in many complex structures that can imitate organisms on the lowest levels in biology. If Conway's game of life is maybe remade into something where neuron connections are made and these neuron connections can interact just like neurons interact in reality, a primitive simulation of a human brain could be formed. Although, yes I agree the concept can use some detail to improve its merit.

[pears]

This topic seems right up EdEarl's street. He provides some links in the following post that may be of interest.

 

http://www.scienceforums.net/topic/79009-brain-vs-computer/?p=771152

 

In particular the Wikipedia article https://en.wikipedia.org/wiki/Artificial_general_intelligence refers to whole brain emulation. I've not looked at them in detail but you may find something useful/informative.

 

If you are interested in replicating consciousness however I agree with iNow. We just don't understand enough about minds or how the brain produces consciousness yet. Also trying to test whether a simulated brain is conscious will be really difficult if not impossible. If a simulation behaves as if it is experiencing things, does that mean it actually is experiencing things?

Edited August 26, 2014 by pears

[Sato]

This is whole brain emulation. There are a few problems with achieving it right now:

 

First of all, we don't exactly know how / why every neural circuit and transmitter and section works, in fact we don't know for the most part. Even if we could see all of these interactions, we'd have to make sense of them in order to remove error from the "brain emulation" or even try to understand what's going on past just a bunch of particle simulations. We could use machine learning techniques to analyze the data from the brain imagers and derive rules and patterns for us, and then maybe we could make a whole brain emulation that we could do something with.

 

Secondly, however, we currently can't get such good images from imagers/scanners. Eg, a typical clinical MRI's voxel (volumetric pixel) resolution is a few milimeters, and I think the highest we've ever achieved is on the scale of micrometers by increasing the magnetic field strength to some extent. In order to capture every individual neurotransmitter necessary for a whole understanding of the interaction, we'd need a spatial resolution of a few nanometers, and though we can do so by placing samples of a brain under a microscope, the voxel depth would also be on the scale of nanometers. The problem here is that the brain is ~1400 cubic cm, and a scanner wouldn't get past ~1^-7 cm, maybe a bit more. Project BigBrain tried to evade this problem by taking very thin slices of a dead human brain and scanning them, but they only achieved a resolution of 20 cubic micrometers, far of from what's necessary to see the actual interactions.

 

Thirdly, however, even if we were able to image with such a depth and resolution, it would take an extremely long time to transfer, process, and store each individual component of each neuron and each neurotransmitter in a computer. Consider that a human brain has on average 80 billion neurons, 10^15 synapses, and active neurotransmitters on a similar scale. Now consider that in order to run the whole brain emulation you'd have to perform a set of physical computations for each each one; if you wanted to track a single thought—a subvocalization, a visualization—it would likely take years, if not past a lifetime. If you wanted to have a brain emulation in real time (that is, as fast as a normal human), well, just as with the imaging technology, we don't have it yet. Lots of complex simulations have been developed and used, such as hydrocode running for supernova simulations, but those track fluids on the scale of cubic meters and still take months to complete on even the most state-of-the-art (non-quantum) supercomputing clusters.

 

Someone who some people I know know, and so has popped up on my radar, is actually working on the problem of whole brain emulation by way of imaging and analyzing the brain of a nematode, which only contains ~300 neurons, which is a step forward.

 

Hope this helped.

[iNow]

In the meantime, here are some additional cool sites:

 

https://www.humanbrainproject.eu/brain-simulation-platform1

https://www.msu.edu/~brains/brains/human/index.html

http://human.brain-map.org/static/brainexplorer

http://scalablebrainatlas.incf.org/main/coronal3d.php?template=LPBA40_on_SRI24&

 

An article discussing exactly this in Nature: http://www.nature.com/news/computer-modelling-brain-in-a-box-1.10066

 

Oh, and a TED talk that seems way up your alley:

 

http://www.ted.com/talks/henry_markram_supercomputing_the_brain_s_secrets

[Unity+]

Thank you everyone for the resources posted. I am assuming the general answer is "It is possible theoretically, however the resources required just have not been met yet."

[iNow]

^ Exactly. Emphasis on the "yet."