Faltering Neuroscientists Improperly Offer Computer Science Papers as Examples of Neuroscience Progress

 A recent scientific paper was entitled "The coming decade of digital brain research: A vision for neuroscience."  Consisting of little more than 100 paragraphs, the paper has more than 100 authors, reminding us of the ridiculous tendency these days for neuroscience papers to have excessive numbers of people listed as authors. We may wonder: what rule was going on here, a rule of "only one paragraph per author"?

Papers like this may remind us of the sad state of current neuroscience, in which it seems that the Supremely Important Things are not good science methodology and strict accuracy of statements but instead an author's paper count (his total of published papers) and an author's citation count (how many citations the papers have got).  So we have endless Questionable Research Practice papers following a bad methodology, often making untrue but interesting-sounding claims in their paper titles or paper abstracts, papers that typically list more than ten authors each. It is as if "quick and dirty" is the operating rule rather than "slow and clean," as if people were trying "at all costs" to increase their count of published papers and the numbers of citations such papers get, and paying relatively little attention to the quality of such papers. It is as if "quantity not quality" is the operating principle. Things get supremely ridiculous when researchers then dishonestly make claims such as "I am the author of 75 published scientific papers" when such a researcher has merely co-authored most of those papers, with the co-authorship mostly merely being a measly "decile co-authorship" in which the author is only one of ten or more listed authors. 

The paper starts out immediately by making a boastful baloney claim, the claim that "in recent years, brain research has indisputably entered a new epoch." No, the kind of brain research we are getting in the 2020's is very little different from the kind of brain research we got in the 2010's.  The authors discuss the Human Brain Project, which (despite billions in funding) failed to make any real progress in supporting the "brains make minds" claims that neuroscientists like to make, completely failing to provide evidence of a brain storage of memories. The paper authors attempt to persuade us otherwise. They make the following statement:  "To give a few examples, research in the project has led to new insights into the mechanisms of learning (Bellec et al., 2020; Cramer et al., 2020; Deperrois et  al., 2022; Göltz et  al., 2021; Jordan et al., 2021; Manninen et al., 2020; Masoli et al., 2021; Stöckl & Maass, 2021; van den Bosch et al., 2022), visuomotor control (Abadía et al., 2021; Pearson et al., 2021), vision (Chen et al., 2020; Svanera et al., 2021; van Vugt et  al., 2018), consciousness (Demertzi et  al., 2019; Lee et al., 2022), sleep (Capone et al., 2019; Le Van Quyen et  al., 2016; Rosanova et  al., 2018), spatial navigation (Bicanski & Burgess, 2018; Northoff et al., 2020; Stoianov et al., 2018; van Beest et al., 2021), predictive coding and perception (Oude Lohuis et al., 2022), as well as language (Dehaene et al., 2015) and has resulted in new theoretical concepts and analysis methods."

The claim that any of these studies provided "new insights into the mechanisms of learning" is incorrect, and neuroscientists lack any understanding of any neural mechanism of learning.  Neuroscientists give us nothing other than empty hand-waving whenever they try to speak of a brain mechanism of learning. Let's take a close look at the papers cited, to see how none of them provide any insights into a brain mechanism of learning:

• "Bellec et al., 2020": This refers to the paper "A solution to the learning dilemma for recurrent networks of spiking neurons" here. This is a computer science paper and mathematics paper that brags about some Atari video game result produced by a software program. It is a not a paper involving any experiments with living organisms or any new observations of living organisms or their cells. 
• "Cramer et al., 2020": This refers to the paper "Control of criticality and computation in spiking neuromorphic networks with plasticity." This is a computer science paper that talks about some result produced using an electronic hardware chip that was inaccurately described as "neuromorphic," a term presumably meaning "like a neuron." A visual of this chip shows something looking nothing like brain tissue.  This is  not a paper involving any experiments with living organisms or any new observations of living organisms or their cells. The visual below shows some misleading labels and captions used in the paper. 
• "Deperrois et  al., 2022": This refers to the paper "Learning cortical representations through perturbed and adversarial dreaming." The paper discusses experiments done with some fancy electronic device or computer software implementation  depicted in Figure 8 of the paper. This is a not a paper involving any experiments with living organisms or any new observations of living organisms. 
• "Göltz et  al., 2021":  this refers to the paper "Fast and energy-efficient neuromorphic deep learning with first-spike times." The paper discusses experiments done with some fancy electronic device or computer software implementation, misleadingly using the term "neurons" repeatedly for parts of such a thing that are not actually neurons. This is  not a paper involving any experiments with living organisms or any new observations of living organisms or their cells. By now we can learn the lesson that whenever you read the word "neuromorphic" in a science paper title (a term meaning "like neurons"), the paper is talking about some computer software and/or computer hardware setup rather than something actually in a brain. 
• "Jordan et al., 2021": this refers to the paper "Evolving interpretable plasticity for spiking networks." The paper discusses experiments done with some fancy electronic device or computer software implementation. This is  not a paper involving any experiments with living organisms or any new observations of living organisms or their cells.
• "Manninen et al., 2020": this refers to the paper "Astrocyte-mediated spike-timing-dependent longterm depression modulates synaptic properties in the developing cortex." This paper involves what it calls "in silico experiments," a term meaning experiments done with some fancy electronic device or computer software implementation. This is  not a paper involving any experiments with living organisms or any new observations of living organisms or their cells. 
• "Masoli et al., 2021": this refers to the paper "Cerebellar golgi cell models predict dendritic processing and mechanisms of synaptic plasticity." The paper discusses experiments done with some fancy electronic device or computer software implementation. This is  not a paper involving any experiments with living organisms or any new observations of living organisms or their cells.

• "Stöckl & Maass, 2021": this refers to the paper "Optimized spiking neurons can classify images with high accuracy through temporal coding with two spikes." This is  not a paper involving any experiments with living organisms or any new observations of living organisms or their cells.
• "van den Bosch et al., 2022": this refers to the paper "Striatal dopamine dissociates methylphenidate effects on value-based versus surprise based reversal learning." Unlike all the papers discussed above, this paper actually involved some experiments with living organisms: some humans who were given two drugs. The experiments report a slight improvement in performance on those given one of the drugs, but only a very slight performance. Supplementary Table 1 in the Supplementary Information  shows those given a placebo scored .90, that those given one drug scored .94 and those given another drug scored .88.  The results are not very impressive, and do not constitute anything like "new insights into the mechanism of learning."

So I have now discussed all of the papers that were claimed to have provided "new insights into the mechanism of learning." by the recent paper "The coming decade of digital brain research: A vision for neuroscience" with 100+ authors. We have been seriously misled. The paper has boasted that the Human Brain Project produced  "new insights into the mechanism of learning," and has given the papers above as its evidence. None of the papers provides any evidence of any such thing as a neural mechanism of learning. All of the papers except one are papers that were mere computer science papers, and not a paper involving any experiments with living organisms or any new observations of living organisms or their cells. Repeatedly in these papers there occurs the deception of referring to purely computer software components or electronic components, and misleadingly calling them "neurons," "synapses," and "dendrites." An example of such a deception is shown below, which is a visual from one of the papers mentioned above:

 

The paper "The coming decade of digital brain research: A vision for neuroscience" with 100+ authors has no description of any progress made in explaining how brains could produce learning, memory, consciousness, self-hood or thinking. The paper fails to use the word "engram." The paper offers some lame excuses for the lack of progress in these areas. It states this: "Our current understanding of the mechanistic operations which subserve cognitive functions, such as memory or decision making, is limited by the scale and precision of existing technologies—simultaneous microscopic recordings are limited to a few brain regions, while full-brain imaging lacks the spatial and/or temporal resolution needed." There are no "mechanistic operations which subserve cognitive functions, such as memory or decision making." The claim that there are such things is a groundless myth of the belief community of neuroscientists.  The excuse given is one that sounds as phony as a three-dollar bill. For decades microscopes have had a power which should have been sufficient to discover memories stored in brains, if they existed.  

Below is a diagram from the paper "Materials Advances Through Aberration-Corrected Electron Microscopy." We see that since the time the genetic code was discovered about 1953, microscopes have grown very many times more powerful. The A on the left stands for an angstrom, a tenth of a nanometer (that is, a ten-billionth of a meter). 

Currently the most powerful microscopes can see things about 1 angstrom in width, which is a tenth of a nanometer. How does this compare to the sizes of the smallest units in brains? Those sizes are below:

Width of a neuron body (soma): about 100 microns (micrometers), which is about 1,000,000 angstroms.

Width of a synapse: about 20-30 nanometers, about 200-300 angstroms.  

Width of a dendritic spine: about 50 to 200 nanometers, about 500 to 2000 angstroms.

Clearly the resolution of the most powerful microscopes is powerful enough to read memories stored in neurons or synapses, if such memories existed. And more than 14,000 brains have been microscopically studied in recent years. The failure to microscopically read any  memories from human brain tissue is a major reason for thinking that brains do not store human memories.  

Besides failing to find specific memories and items of learned knowledge by microscopically examining brains (such as the information that the New York Yankees belong to the American League of US baseball), scientists can find no evidence of a mechanism for storing learned information in brains.  If such a mechanism existed, its fingerprints would be all over the place. Since humans can learn and remember so many different types of things (sights, sounds, feelings, facts, beliefs, opinions, numbers, smells, tastes, physical pains, physical pleasures, music, quotations, and so forth), any brain mechanism for storing all of these things would have a massive footprint in the brain and in the genome. No sign of any such thing can be found. The workhorses that get things done in the body are proteins, and humans have more than 20,000 types of proteins. No one has ever identified a protein that helps to write a memory of experiences or numbers or words to the brain or neural tissue, in any kind of way that helps explain how memories or knowledge could be stored in brains.  Of course, you can find studies maybe showing that protein XYZ was used when someone learned something, but that does nothing to show a mechanism of memory storage. 

The paper "The coming decade of digital brain research: A vision for neuroscience" with 100+ authors fails to seriously discuss the gigantic rot at the core of today's neuroscience: the massive production of irreproducible results caused by countless experimenters doing "quick and dirty" research following Questionable Research Practices such as inadequate sample sizes, lack of blinding protocols, and the use of poor measurement techniques such as judgments of freezing behavior.

The paper makes this laughable statement: "Brain simulation is expected to play a key role in elucidating essential aspects of brain processes (by demonstrating the capacity to reproduce them in silico),

such as decision-making, sensorimotor integration, memory formation, etc." What silly drivel this is. Computers are computers, and brains are brains. You do nothing to show that brains produce decisions or that brains form memories by showing that computers can make decisions or that computers can store new information in computer memory systems. 

It is rather clear what is going on. Our neuroscientists are getting nowhere trying to show that brains can produce thinking or learning or store memories or retrieve memories (things that brains do not actually do).  To try to cover up their lack of progress, neuroscientists are appealing to computer science papers that merely show examples of computers storing, retrieving and processing data. It's kind of like a husband who is failing to earn a salary, and bragging to his wife about all the virtual money he is making in some life-simulation game such as The Sims. 

Click to see left column more clearly

Click to see left column more clearly