Sunday, May 02, 2021

AI research may have hit a dead end

I found a link to a very interesting article titled "Artificial intelligence research may have hit a dead end" followed by the comment "Misfired" neurons might be a brain feature, not a bug — and that's something AI research can't take into account" (see this).

Also Philip K. Dick's 1968 sci-fi novel, "Do Androids Dream of Electric Sheep?" is mentioned.  Would an intelligent robot  (if it were still a robot) dream?

AI models the brain as a deterministic computer. Computer does not dream:  it does just what  is needed to solve a highly specialized problem (just what a top  specialist does in his job;  computer is the  idol of every professional highflier). 

Computerism assumes  physicalism denying such things as genuine free will but this is not seen as a problem.  Also the mainstream  neuroscientist believes in physicalism. Some computational imperialists   even claim that physics reduces to computerism.

1. Is 95 per cent of brain activity mere noise?

What might be called neuroscience of fluctuations has however  led to a strange conclusion: 95 per cent of brain's activity and therefore metabolic energy seems to be used  to  generate fluctuations,  which in standard neuroscience represents noise.   Neuroscientists have routinely averaged out this "noise" and concentrated on the study of  what can be regarded as  motor actions and sensory input. These contributions seem to represent only ripples in a vast sea of activity.

[Amusingly, junk DNA corresponds to 95 per cent of DNA in the case of humans, as the article observes.]

By the way,  EEG is  still often regarded  as a mere noise. This  represents a similar puzzle: why the brain would use a lot of metabolic energy to send information to outer space: coding of information about contents of consciousness and brain state indeed requires a lot of metabolic energy.   To sum up, the brain seems to be diametrically opposite to a computer in the sense that spontaneous fluctuations are poison for a computer but food for the brain.

What article suggests  is that this 95 per cent could correspond to "dreaming"  that is  imagination. Ability to imagine  would give rise to intelligence rather than the property of being a dead automaton. Dreams would be freely associating cognitive fluctuations - whatever that might mean physically. Interestingly, it is mentioned that newborns dream twice as much as adults: they must learn. One can learn by imaging, not merely by doing all possible mistakes in the real world.

What can one say about these findings in the TGD framework?

2. Could fluctuations be induced by quantum fluctuations in quantum critical Universe of TGD?

Consider first the TGD interpretation of quantum fluctuations.

  1.  TGD Universe is  quantal  in all scales. Zero energy ontology (ZEO) allows to overcome the basic objection that the universe looks classical in long scales: ZEO view about quantum jumps forces the Universe to look classical for  the outsider. The experiments of Minev et al indeed demonstrated this concretely.
  2. TGD Universe is also quantum critical in all scales. Quantum criticality means that the system is maximally complex  and sensitive  for perturbations. Complexity means that the system is ideal for representing the  external world via sensory inputs. By criticality implying maximal sensitivity it is also an ideal sensory receptor  and motor instrument.
  3. The basic characteristic of criticality are long range fluctuations. They are not random noise but highly correlated.  Could the fluctuation in  the brain correspond to quantum fluctuations. 
 Long range quantum fluctuations are not possible for the ordinary value of  Planck constant.  
  1. Number theoretical view about TGD, generalizing ordinary physics of sensory experience to the physics of both sensory experience and cognition, leads to the prediction that there is infinite hierarchy of phases of ordinary matter  identifiable as dark matter and labelled by the values of effective Planck constant heff= nh0, n is dimension for an extension of rationals defined by a polynomial determining space-time region.
  2. The value of n serves as a measure for complexity and therefore defines a kind of IQ. The longer the scale of quantum fluctuations, the higher the value of n, and the larger the heff, and the longer the scale of quantum coherence. Fluctuations would make  the brain intelligent. Their  absence would make the brain a complete idiot - an ideal computer.
  3. The higher the value of heff, the larger the energy of  the particle when other parameters are kept as constant. This means that intelligence requires metabolic energy feed to increase heff and keep its values the same, since heff tends to be spontaneously reduced.
One can however argue that since the brain consists of ordinary matter,  brain fluctuations cannot be quantal. 
  1. In TGD they would be induced by quantum fluctuations at the level of  the magnetic body (MB) having a hierarchical onion-like structure. The dark matter would be ordinary particles with heff=nh0 at MB and since heff/h0 serves as a measure of IQ it would be higher for dark matter than for ordinary biomatter. MB containing dark matter would be the "boss" controlling the biological body (BB).  
  2. The quantum coherence of MB would force ordinary coherence of ordinary biomatter as a forced coherence. Ordinary matter would be like soldiers  obeying the  orders and in this manner behaving apparently like a larger coherent unit.
MB would receive sensory input from BB and control it by using EEG realizes as dark photons. This would explain EEG and its probably existing scaled  variants.

3. TGD view about sensory perception, motor actions, and dreaming and imagination

The proposal of the article was that most of the brain activity goes to "dreaming". Dreaming, hallucinations,  and imagination are poorly understood notions in neuroscience.  TGD provides a rather detailed view about these notions.

  1. What  distinguishes TGD from neuroscience is that sensory receptors are assumed to serve as carriers of sensory percepts. Zero energy ontology (ZEO)  providing a new view about time and memory makes it possible  to solve the basic objections related to  the phantom limb phenomenon: pain in  the phantom limb would be sensory memory.   
  2. The assumption that sensory percepts are artworks rather than passive records  of sensory input   requires virtual sensory input from brain to sensory organs and build-up of the final  percept by  pattern recognition -  an iterative procedure involving very many forth-and back signals. Nerve pulse transmission is quite too slow a process to allow this and signals propagating with maximal signal velocity are suggestive.
  3.  Nerve pulses and neurotransmitters  would not represent real communication but give rise to  temporary intra-brain communication lines along which communications as dark photon signals would take place with maximal signal velocity using dark photons  (characterized by heff/h0=n) transforming to biophotons in an  energy conserving manner. 

    Neurotransmitters and also other  information molecules (hormones, messengers) attached to receptors would serve as bridges fusing  permanent but disjoint communication lines along axons to a connected temporary communication line for dark photons to propagate.  Nerve pulses would also generate generalized Josephson radiation allowing communications   between biological body (BB) and magnetic body (MB) using EEG.  Meridian system could be a permanently connected system of communication lines.

    This picture leads to a concrete  proposal about the roles of  DMT and pineal gland  concerning imagination and dreams and hallucinations. 

Returning to the original topic,  the natural question is following: How large fraction  of the 95 percent of brain activity goes to  feedback not present in the brain of the standard neuroscience? This would include  the construction of the feedback to sensory organs as  virtual sensory inputs to build standardized mental images. Dreams are a special case of this. There  is  also the virtual sensory input which does not reach sensory organs and gives rise to imagination, in particular internal speech.

 Similar picture applies to virtual  motor input and the construction of motor output as "standardized motor patterns" - this notion makes sense only in ZEO.  Note that the feedback loop could extend from brain to MB. 

There is also an interesting finding related to motor activities. In the experiments made for rats it is found that the spontaneous brain activity increases dramatically as the rat moves. This brings in mind a lecturer who moves forth and back  as he talks. This rhythmic motion could give rise to a brain/body rhythm  coupling the lecturer to a layer of MB with large heff. Its   quantum coherence of MB would induce ordinary coherence of BB in body scale and with large heff and raise the "IQ" of the lecture.  Thinking requires motion!

For a summary of earlier postings see Latest progress in TGD.

Articles and other material related to TGD. 

No comments: