Notes on neuromodeling

The quest to build consciousness

I came across a very interesting modeling initiative on human brain, the Human Brain Project, that gets a billion € worth of funding from the EU- see http://www.humanbrainproject.eu/ for yourself, and also the brilliant Wired article that alerted me to the topic. Commenters (laymen and professional alike) have already pointed out many caveats to the success of modeling a whole working human brain, such as the enormous complexity and the fact that brains have bodies attached which modifies their working significantly. Well - maybe computing power will overcome the complexity, and maybe they will simulate a working body as an extension to the simulated brain. What bugs me with this approach is, why build a whole brain simulation top down? It will be a sort of an average of all the brain research data ever accumulated. Wouldn't it be like trying to create a living tree out of sawdust?
Why not plant a seed instead?

Scale model 1:10^10

Attempts to a simulation of anything, I assume, always come with the scale problem attached. What level of detail should we simulate? The HBP announces it will simulate many levels. If you want a working, interactive simulation, the lowest level should always be the only level that's needed - right? The "lowest level" here means the smallest detail of simulation that can possibly, realistically, be reached. That would not be the level of atoms, I suppose. Not molecules either. Single cells perhaps? Single neuron simulation itself is costly: an IBM Blue Gene/P supercomputer simulates 10'000 neurons with 1 billion synapses 300 x slower than real time (Markram's previous project, see http://bluebrain.epfl.ch/). Now, human brain has 86 *billion* neurons. At that speed, it will take Blue Gene some 800+ yrs to simulate one second of a whole working brain. No doubt, computing speed will go up in following decades and perhaps quantum computing will help the modelers in their feat. Faster, we want faster! How can you have a discussion with an artificial intelligence that responds "Hello Dave" in the afternoon when you have greeted it in the morning?

Franken-mind

The details problem is one, the interactive realism is another. Will a sawdust-tree equivalent of a brain truly simulate a healthy, working human brain? If it speaks, we will find out of course. But what language will it speak since it's put together from bits and pieces of neuroscientific research findings from all over the world? It needs to learn. Learning a language takes time and practice, and requires interaction with the environment. As does any other learning process, and any emotional and intellectual development. Basically we would try to teach the world to a grown-up brain that has no continuous experience of the world at all - only little bits of pieces of shards of fragments of memories. In other words - a completely shattered mind; probably worse than that caused by any known brain disease. The "we were made like this" approach clearly is not the way here. Instead look at the nature. Most complex organisms begin at a single cell and develop in interaction with the environment.

Creation vs. evolution

The development of an individual (human or other organism) is not evolution, but a product of it. Certain limits apply: we do not develop wings or tail - or eletro-shock organs - but do learn a language or two, grow tall or not, become athletes and get depression, depending on our surroundings. Similarly, a simulated brain should start at embryo-stage or earlier so it would develop in interaction with the world. And, to avoid the  scale problem, the world should not be simulated but real, the same physical reality we inhabit. Then - with the future speed-up of computing equipment - we could simulate whole lives and get information on what factors have effect on brain development. These artificial minds would share our experience and mindset. There would be no reason to fear the science fiction nightmare of a super-powerful computer mind that aspires for world domination and enslavement/destruction of humankind. No fear of Skynet, HAL9000 or Matrix.

Existence in time

1. The puzzling issue about the past. Time, as understood by Einstein (as I understand it), is a dimension along which energy is spread. Unlike spatial dimensions, it only allows movement in one direction. If objects could travel in the other direction, considered "the past" by us, what would await them there? We are in the present, out of the past, and so is everything else around us. Is there still something left back there? A copy of the universe? A past version of all the energy and matter? If so, we need to ask: what is the "refresh rate" of the universe? If the 4-dimensional universe consists of a series of frozen snapshots, arranged along the time dimension, how many of those snapshots do we need for e.g. one second of existence? Or is existence truly continuous, so that no matter how small time interval you observe, you can always still find change in the state of the universe during that interval?

2. ...and the present
Also, I get it the theory of relativity (the general version) states that there is no fixed point of "now". Time moves slower with increased velocity. So my present becomes the future for a jet pilot, although the time lag is extremely small and can never be utilized. Still, objects with different velocities (like the jet pilot and me) co-exist despite the time lag. They share the same universe and can share information with each other. I assume that this co-existence in time equals the concept of "now" as we understand it, in relation to past, which we have no way of reaching or future, which we will reach only when a given amount of time has passed - not earlier or sooner.

3. The future?
As mentioned in 2., we can't reach the future - our relative future - earlier than what the relative passage of time permits us to. This implies, to my mind, that the future is fixed, set and waiting for us. It exists in the sense that the past exists, as is assumed here in 1. It exists immutable, connected to the events in the relative past but not dependent upon it.

4. A frozen, infinitely massive universe. What could be thought of this (1., 2. and 3.), is that time-space is a possibly infinite continuum of mass and movement (i.e., energy) that forms a series of events of logical and natural order; and, that causality or choice do not exist in the reality of this time-space continuum. The energy of the time-space (its mass) is the total of this continuum. If there is no "refresh rate" as in 1., i.e. if the time-space is truly continuous and has no gaps in the time dimension, it's mass would be infinite even if the universe had a fixed end point in time. According to current knowledge, however, there is no such end point to be reached - universe will just keep expanding infinitely. This leads to it having infinite mass even in the case time was quantized, so that it could only be divided so far and not further - a "smallest unit" of time existing. Even in this case the fact that the universe expands forever would make its total mass infinite, because the space continuum would be infinitely "long", or rather, of infinite duration.