Intelligent neurons

Postings to particle physics blogs seem to have degenerated into boring n-sigma talk, which faitfully reflects the deep stagnation in particle physics theory. In biology and neuroscience situation is totally different. For instance, this morning I found a highly interesting article about about intelligent neurons.

The article gives a link to a Youtube video about a robot, which can move and has learned to avoid obstacles. The robot is controlled by neurons from rat embryo, which have grown to form a densely connected population of about 100.000 neurons. Some kind of training is is performed.

The neuron population has developed a sense of vision. We do not know whether the population sees as we do. Also we can "see" using other senses such as tactile sense and this must be based on building a geometric symbolic representation from environment using the sensory input. Neuron population utilizes the visual input in its motor activities made possible by the coupling to a simple robot and learns to avoid obstacles.

The interpretation in the framework of standard neuroscience is challenging. Neuron population behaves as if it were a conscious creature with intentions, goals, ability to see, and ability to move in a manner taking into account the constraints posed by the environment. But a neuroscientist who has read his text books is materialist and not willing to admit that neuron population could behave like a conscious creature. He tries to understand the situation on basis of classical computation realizing some kind of unconscious general intelligence and realying on some mystical algorithm allowing to recognize patterns and adapt to almost any situation. In principle the neuron population should be replaceable by a general purpose computer program if computationalism works. This kind of extreme flexibility means that this program must be very very long, perhaps too long to be fast enough and be realizable using finite metabolic and material resources!

In the second experiment discussed in the posting ferret's brain was rewired. Visual sensory input from optic nerve was redirected to the brain regions, where the input from ears is normally processed. Despite this the brain region automatically re-configured to make sense of the visual data and allowed the ferret to see with 1/3 of the normal vision. The structures presumably needed for seeing were automatically formed in the brain section used to process sound.

1. The experiment excludes the idea that genes are responsible for the differences between auditory and visual regions of brain. Neurons are same neurons everywhere in brain and it is self-organization, which is responsible for the specialization to see or hear - or rather, to process sensory inputs whatever they are. Top-down instructions at brain level can be excluded since the ferret's brain does not know about rewiring. Also the feedback teaching the neurons to do the "right" thing as in the first experiments was absent. It seems that the neuron population behaves as a conscious creature able to self-organize to utilize the input from sensory organs, be it visual or auditory.

2. In the ordinary neuro-science this seems to leave only one option. If neurons are assumed to be un-conscious, the rewiring between neurons should somehow give rise to a symbolic representation about the geometry of the environment. The great mystery is how a mere wiring could give rise to various qualia such has color even in the normal situation. The idea that different wiring topologies could give rise to different qualia looks utterly implausible.

3. The symbolic representation of the environment could use only geometric data. Whether it involves also qualia such as colors could be clarified an experiment using human subjects and this raises ethical and probably also practical issues. One could however test whether ferret learns to recognize objects with different color: for instance, green object could course a pleasant sensation and red object an unpleasant sensation. A variant of this test might be possible even in the case of neuronal robot.

What about the interpretation in TGD framework? Both experiments conform with the general TGD inspired vision about brain.

1. For the simplest option the fundamental sensory qualia reside at the level of sensory receptors. This is not the only option but it is very attractive one. I do not repeat here the arguments allowing to circumvent the objections against qualia at the sensory receptors: phantom limb, etc... The sensation of color involves quantum entanglement of magnetic body, brain, and retina. In the latter experiments auditory regions would entangle with retina and the prediction is that also now color qualia are present.

2. Brain builds up symbolic representations by decomposing the sensory input to objects and giving them names. Sensory organs take care of the "coloring" of the resulting sensory map. Sensory feedback from neurons is needed since our perceptions do not represent what is there but a caricature exaggerating the important features. Sensory feedback is in terms of dark photons from neurons to retina with non-standard value of Planck constant. We call them bio-photons when they leak out by transforming to ordinary photons.

3. Neuronal lipid layer serves as the analog of computer monitor screen with each lipid carrying various basic attributes associated with neuronal sensory experience, which remains unconscious to us. This explains grandma neurons able to recognize some particular sensory input. Also neurons possess various primary qualia but sensory organs have specialized to produce sensory qualia at our level of the self hierarchy.

4. Neurons are conscious creatures able to co-operate because they have a collective magnetic body controlling the neuron population, and can therefore rapidly adapt in changing environment as in the first experiment. The presence of the magnetic body of course distinguishes sharply between TGD inspired and neuroscience explanations.