Scanned Brain Simulation

iMmortality

Rather trying to design a brain thru AI; why not just do a detailed scan of a human brain, turn it into a neuron model, and then just "run" it?

I realize our scanning technology is probably not there yet, but we could start testing such on bugs, birds, etc.

One approach they used for other purposes was to freeze a body (terminal subject volunteered before-hand), then slice off a millimeter or so at a time with lasers, and then image the remaining section. For a brain scan, the slices would have to be far smaller than a millimeter, and I don't know if the chemical states of each cell can be determined this way. It may take multi-spectrum imaging.

I remember reading about something similar to this somewhere, but never saw a follow-up.

I fear YouCantUnderstandaProcessByStoppingIt. And anyway you will need some input and output too, so scanning the brain alone simply will not do.

The input and output does not have to be sophisticated. If it is truly a working model of the brain, it will eventually learn to communicate via whatever I/O paths given to it. It may even be Morse Code-like at first. I don't think it would be too hard to project written instructions via the eye connections. Audio and image input is fairly-well understood compared to other aspects of the brain.

The brain can learn only by relating new things to known things. It might be somewhat different with infants or even embryos. But then what would you get? A pocket calculator? (being sarcastic)

''And vision/audio is not all input there is. I bet that the simulated brain would quickly stop doing anything sensible if it isn't fed input from the lower parts (i.e. body signals and chemical feedbacks).

This is always a good thought experiment. The scanning technology would have to be far more advanced that what we have today. In particular, it would not be sufficient to just scan in a graph of connections between neurons, because we now know that glial cells are critical to how the brain works.

They appear to be mostly about long-term maintenance and cleanup. It is unclear how much effect they will have on our model since it is not subject to physical degeneration.

So if you take it to the next level, we would scan every cell and have knowledge of its "connections". That too would be insufficient, because the cells don't all work the same, and the cells have information content stored internally, not just through connections. The next level, scanning each atom, is even further from our reach today, but it is also likely to be insufficient. Currently, scanning something as large as a human brain, would take quite a while. As someone mentioned above, there is also information content in the running state of the brain - the kind of state that would have to stop in order for scanning to take place. So let's say we invent a machine that can scan each atom in a human brain instantaneously, without disturbing any state.

I am still not understanding why this is allegedly important. Temporary confusion should dissipate, like waking up from a confusing dream. A confusing dream does not cause us to crash. As far as "each cell being different", the model has to only be sufficient, not perfect. The important thing is that differences or internal info be parametrized in some way. We don't know how many parameters is "good enough" at this point. And perhaps the power and/or speed of digital can be used to compensate for the lossy-ness of the model. -t

What we have then is not so much a brain simulator or a human body simulator - it is a universe simulator. The problem there is that we don't know how the universe works quite yet. The human brain may only work in the presence of certain quantum effects. So we may take a guess at universe simulation.

There is no evidence that quantum effects play a significant part (other than "implementing" the existing physical and chemical models). We don't have to be perfect. After all, we damage our brains with alcohol, diabetes, caffeine, etc., all the time, and it still functions around 95% despite the damage. And people have been poisoned, severely crippled, etc. and live on with a still-functioning brain. Thus, we don't need a "perfect" simulation. That is the nice thing about biological designs: they accept a much heavier amount of "distortion" without crashing like our digital stuff does.

There is actually some evidence that neural systems of some animals do indeed take advantage of quantum effects. But none of it seems beyond (sufficient) simulation at this point. It just complicates the model, not ends it.

If we managed to get universe simulation close enough to right, and still have the instant scan machine, then there are further problems with input. There is evidence that humans enter a comatose state when deprived enough sensory input for enough time. Your assumption about simple MorseCode-style communication is incompatible with that evidence. There is also the question of what kinds of signals are "input" to the brain. Things like blood pressure, for example, change the physical structure of the brain continuously. We don't know if or how those background inputs affect cognition.

I agree that it will probably take some experimentation to find out if those kinds of things make a sufficient difference. If you wake up one day like Helen Keller, it is not the end of the world. You will still learn to communicate and cognitive ability is still there.

None of that makes the idea impossible - they are just some obstacles and potential obstacles to think about. There is a small but inactive online community for this. This idea has come to be known as "mind uploading", and you can find some other resources by searching google for that phrase. The first result, http://www.ibiblio.org/jstrout/uploading/MUHomePage.html, has a good high-level look at it. -- MichaelSparks

The Millenium Simulation might have simulated the entire universe but the problem is mathematically not so complicated and is primarily a number-crunching problem. Simulating a human brain is very different from this "simple" equation solving. Brain input from the body is needed from both chemical (eg hormones) and nervous pathways. Both are essential for the production of self-consciousness (which I consider necessary for simulating a human brain). Cutting off the bodily signal pathways has a severe impact on consciousness, self-awareness, emotional functioning etc. What is often left out is the fact that not only the nervous system is essential but also the chemical communication and regulatory mechanisms between the body and the brain via hormones etc. A brain alone does not make a human being. Brains haven't developed as an independent unit but as part of a bigger unit with manipulators, sensors, sewage system etc. Antonio Damasio has written three pretty readable books on the workings of the brain, the mechanisms of how emotions and feelings are produced and gives many examples by telling the cases of several of his neurological patients - See Antonio Damasio's page on the USC college website for more info: http://www.usc.edu/schools/college/bci/faculty1008328.html -- Albedo

Yes, but it still may be possible to work around these other connections. Do we really need a tight relationship to our poop system to be conscience? (However, some of you indeed act like you are heavily connected to your anus.) Unlike a biological system, experimentation to "get it right" is generally easier with a simulation. You can always re-load the prior state if things go sour. --top