No, That Dead Tissue Study Didn't Show That Neurons or Dendrites Compute

When you read news relating to neuroscience, you will typically read a great deal of baloney and shameless hype. The first reason for this has to do with the runaway hype and exaggeration that is currently going on in regard to Internet sites reporting scientific research. Major websites have learned the following fundamental formula:

                                     Clicks= Cash Income

The reason for this is that major websites make money from online ads. So the more people click on a link to some science story, the more money the website makes. This means that science reporting sites have a tremendous financial incentive to hype and exaggerate science stories. If they have a link saying, “Borderline results from new neuron study,” they may make only five dollars from that story. But if they have a story saying, “Astonishing breakthrough unveils the brain secret of memory,” they may make five hundred dollars from that story. With such a situation, it is no wonder that the hyping and exaggeration of scientific research is at epidemic levels.

Part of the problem is university press offices, which nowadays are shameless in exaggerating the importance of research done at their university. A scientific paper reached the following conclusions, indicating a huge hype and exaggeration crisis both among the authors of scientific papers and the media that reports on such papers:

“Thirty-four percent of academic studies and 48% of media articles used language that reviewers considered too strong for their strength of causal inference....Fifty-eight percent of media articles were found to have inaccurately reported the question, results, intervention, or population of the academic study”

Another giant reason for the baloney and hype in today's neuroscience news is that neuroscientists simply aren't finding results that back up dogmatic neuroscientist claims about the brain. Specifically:

• Neuroscientists are not finding any real evidence of any system of encoding by which human conceptual knowledge or episodic memories could be stored as neuron states or synapse states (nor do they even have any detailed theory to explain such a thing).
• Neuroscientists are not finding any legitimate evidence of specific learned knowledge stored anywhere in brains (such as learned knowledge that can be read from the brains of dead people or dead animals).
• Neuroscientists are not finding any evidence that could explain how a brain could think or imagine or create abstract ideas.
• Neuroscientists are not finding any evidence of how memories could be stored for decades, and their research is suggesting exactly the opposite, that brains are totally unsuitable for storing information for any length of time longer than a year (the average life of synapse proteins being only a few weeks).
• Neuroscientists are not finding any real evidence that brains look different during thinking or recall than during mental inactivity (brain scans showing no difference of more than half a percent). 
• Neuroscientists are not finding any brain action that could explain the instantaneous formation of memories, which can form so much quicker than can be explained through "synapse strengthening" that would take quite some time.

So what do our neuroscience news sources do when they don't have the kind of research findings they would like to have? They hype like crazy all kinds of dubious stuff. They take weak or faulty or second-rate studies, and try to make them sound like “eureka” breakthroughs.

An example of this can be found in recent stories in the science news press claiming that some evidence of computational ability had been found in neuron or dendrites. The claim is false, and is based on research that found no robust evidence of such a thing. One of the stories (in the online Quanta magazine) had the title “Hidden Computational Power Found in the Arms of Neurons.” The site Singularity Hub had a bunk headline of "Scientists Discovered ‘Mini-Computers’ in Human Neurons—and That’s Great News for AI." No such thing was actually found. The underlying study was able to claim to find the barest trace of “computational power” only after taking dead brain tissue, artificially zapping it with electricity, and then plugging the resulting data into a computer model, which we can assume was carefully programmed to allow such a report of "signs of computation" to be made. As I will explain, someone could use similar techniques to report “computational power” coming from any of many things as witless and non-computing as a stone.

The study (behind a paywall) was the Science study “Dendritic action potentials and computation in human layer 2/3 cortical neurons.” The two main types of biological studies are in vivo studies (using an actual living organism), and in vitro studies (merely using some dead specimen, or some tissue or cells detached from an organism). The “dendritic action potentials” study was an in vitro study. The study used brain tissue that had been removed from human brains during brain surgeries conducted for other purposes. There is a big reason why it makes no sense to be making any claim about a computational ability of neurons from any type of in vitro study using detached human brain tissue.

The reason is that in an actual living brain, every neuron is connected to many other neurons, and is constantly receiving signals from all of those other neurons. It has been estimated that a typical neuron is connected to thousands of other neurons. Each of these neurons is constantly emitting electrical/chemical signals called action potentials. Although estimates of neuron firing rates vary, it is typically estimated that each neuron sends out an action potential at an average rate of once per second or more. All of this creates the most gigantic noise issue, which should prevent reliable computation from occurring in a brain.

We may compare a neuron to a person on a stage in a very large theater or auditorium, at a time when thousands of people in the audience are all shouting at him. Just as it would be impossible for such a person to hear a full sentence from even one of these voices, it should be impossible for any neuron to have the type of noise exclusion  needed for effective computation, when each neuron is being flooded with signals from hundreds or thousands of other neurons. Neurons and synapses are flooded with so much noise in such high amounts that they cannot be doing the things our neuroscientists like to hope they are doing.  Neuroscientists  pretty much ignore this gigantic "neural noise" problem, and turn a deaf ear to what nature is telling us in this regard. 

If the whole audience was shouting at you, you'd be like a neuron

But imagine you take some brain tissue out of the brain, and test that artificially, in vitro. Then by zapping a neuron with a single electrode, you can get a nice clean situation in which all of those signal bombardments are eliminated. But such a situation tells us nothing about a real human brain. For it bears no resemblance to what is going in a human brain in which each neuron is bombarded by electrical signals from thousands of other neurons.

It is just such an in vitro situation that was used by the “dendritic action potentials” study, which got some data from zapping dead brain cells rather than data from living human brains. The data they got did not by itself indicate that anything like computation was going on in the dead brain tissue that was tested.

So the experimenters had to insert their data into something called a computer model. A computer model is some computer program designed to simulate physical reality. Of course, when we are dealing with the near-infinite complexities of human biology, trying to simulate things with computer models is pretty much a joke. You can get pretty much anything you want to show from a computer model. You just program things to favor the desired result.

The study authors claim to have got a little evidence of computation, but merely some minimal couldn't-be-simpler computation called an  “XOR” computation. They claim to have got this after they took their data from zapping dead brain tissue, and plugged it into their computer program. I may note that you could use such an approach to yield such “evidence of computation” from very many things which we regard as being as mindless and non-computing as a stone.

I will give an example. Below (according to the wikipedia article) is the truth table describing the simple-as-it-gets “XOR” computation:

Input		Output
A	B
0	0	0
0	1	1
1	0	1
1	1	0

I could write a computer program to simulate coin flips. There would be four possibilities: (1) a non-successful flip because the coin went down the flipper's shirt; (2) a “heads” flip; (3) a “tails” flip; (4) the coin landing on its edge, which can be considered both “heads” and “tails” (since both sides of the coin can be seen). So now the truth table for this coin flipping would look like the table below, and would resemble the table for the "XOR" computation:

Input		Output (1= successful flip)	Comment
“Heads”	“Tails”
0	0	0	Coin lands in flipper's shirt
0	1	1	Coins lands “tails”
1	0	1	Coin lands “heads”
1	1	0	Coin lands on edge

Using such assumptions, I could simulate the flipping of coins in my computer program. And what would my program suggest? Not very much, but you might claim that such a program suggests that mindless coins can compute. For in such a program, the mindless coin flipping is producing something rather like an “XOR” operation (if we use the analysis above).

Of course, such a program would be no real reason for thinking that a coin can to any degree compute or think. And it would not give us any reason for thinking that big sacks of coins are computing or thinking, or that a whole bunch of coins jiggling around in some machine would be producing anything like computation or thinking. Similarly, studies done with dead brain tissue hooked up to electrodes  (studies relying on plugging the data into a computer model designed to extract some sign of computation) don't prove that living neurons or dendrites can compute.

Just as I have shown how you can do a computer model that seems to yield a little "XOR computation" from simulated coin flips, I could have shown how other computer models might seem to show "XOR" computation from numerous other inanimate mindless non-computing objects. The "dendritic action potentials" paper has produced no robust evidence of computation in the brain. Because it used dead tissue rather than live brains, the “dendritic action potentials” paper was rather like a study trying to prove that gorillas can play baseball, not by doing experiments with live gorillas, but by taking a baseball glove and sticking it on the hand of a dead gorilla, to see if you could pitch a ball into the glove of the dead gorilla without the ball falling out.  

In reality, the human brain bears virtually no resemblance to a computer, and there is still no good evidence that the brain does anything like computation. Brains don't have any of seven major things that computers use to store and retrieve information. Computers compute by using software, and nothing like software exists in the brain. A non-metallic device without software doesn't seem like something that can be doing much in the way of computation.  Someone who claims that a brain without software is "just like a computer" is like someone claiming that a wagon without an engine is "just like a car."