Science fiction is full of examples of artificial minds. From Robbie the Robot from Forbidden Planet and Hal, from 2001, (technically, just a computer) all the way to Battlestar Galactica’s Cylons including the hideous (not!) Caprica 6. One of the best known types of robots are the ones imagined by Isaac Asimov. He was one of the first writers that in addition of creating robot characters, also incorporated in many of his stories a mechanism for creating artificial intelligence. He did this by “inventing” a still fictional technology for creating something called positronic brains. We will not go over positronic brains here though; maybe at some other time.
Regardless of the actual physical form that an artificial intelligence may have, the important thing (from the perspective of this post) is that some kind of mind/intelligence is at play. An implicit question is immediately apparent. How complex does a physical system has to be for the emergence of a human-style mind? In other words, what is the minimal number of interacting neurons that we need for consciousness to appear?
When we humans think about numbers, especially if we need to count a big amount of anything, we tend to think in groups of ten. This is most likely a consequence of (usually) having ten fingers. Because of that, people like working with “nice round numbers” usually some power of 10.
What are these “powers of 10”?
Well, you know: 10, 100, 1000, 10000, etc. We also use expressions like “order of magnitude”, which means essentially the same thing. This is, when something differs from anything else by (usually) a factor of ten, we are talking about an order of magnitude. Similarly, two orders of magnitude is a factor of 100 and so on. We just happen to feel comfortable thinking in those terms.
Now, when thinking about nature with that frame of mind, we come across to a lot of very curious coincidences. For example, the best available data suggests that there are close to 100 (maybe 200) billion galaxies in the known universe, each of them with an average of 100 billion stars. A billion, by the way, is a very big number. To avoid any confusion, what we mean by a billion here is 1,000,000,000, and it is a big number indeed. Just to give you some perspective, if you were to count to a billion at roughly one number per second, nonstop, you would need about 32 years.
Another famous “100 billion” amount is the estimated number of nerve cells in a typical human brain. Oddly, nobody knows exactly where this estimate came from, even though is THE figure cited in virtually every book, every newspiece, etc. Because of that, a research group stepped up to the plate and decided to do the experiments to actually try and count the neurons in a human brain.
The result of that was the following paper:
Azevedo FA, Carvalho LR, Grinberg LT, Farfel JM, Ferretti RE, Leite RE, Jacob Filho W, Lent R, Herculano-Houzel S (2009) Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain. J Comp Neurol. 513(5):532-41.
The leader of this group is an accomplished neuroscientist, Dr. Suzana Herculano-Houzel, of the Federal University of Rio de Janeiro, Brazil. She kindly sent me a copy of her paper as well as many other papers on her research about the evolution of the brain. The paper is very well-written and their conclusions are logical and sound. However, these results triggered a minor controversy regarding the exact number of neurons in a human brain. The controversy was in no way the fault of any of the researchers; rather, I blame the current editorial trend of “byte-size” science that merely throws a flashy headline to get attention. Nothing wrong with that, mind you, except when the headline is not followed by adequate reporting. When this follow up does not happen, it will create deep misunderstandings.
Not so long ago, James Randerson wrote a piece for “The Guardian” titled: “How many neurons make a human brain? Billions fewer than we thought”, based on the Azevedo paper and an interview with Dr. Herculano-Houzel. Randerson’s piece is found here:
Basically, this piece “reports” that the Azevedo paper mainly found that the average number of neurons in humans is close to 86 billion. From the perspective of just this bare fact, yes indeed, we seem to be missing 14 billion neurons. Don’t panic though!
Azevedo’s paper describes how the measurements were done by determining the number of neurons by an interesting cell counting method that I will not discuss here (but it has been described as “brain soup” just in case you are interested in finding out more).
I want to clarify that I am not criticizing the paper. In fact, if find it quite interesting, that is why I read it!
But, back to topic…
There are several factors that immediately come to mind that need to be considered before using this sole paper as a reason to begin rewriting the textbooks to read “The human brain is composed of roughly 86 billion neurons…”
**Brains tend to shrink with age; moreover, the rate of shrinking is surely affected by genetics and environmental factors.
**The sample size was only 4 brains, all of them men’s with an age range from 50 to 71 years old.
**General health and levels of physical activity.
And we could keep going, but you basically get the idea, namely that as it is right now, this data is not sufficient to state that the 86 billion figure is the definitive one.
Other scientists have commented on these issues and have urged a more cautious approach. One of these scientists is a collaborator of the Herculano-Houzel group, Dr. Roberto Lent, also a coauthor of the paper. For more information on Dr. Lent’s point of view go here.
The Azevedo paper reports finding and average of 86.1 +\- 8.1 billion neurons. This means that the actual number of neurons (the approximate range) can be as little as 78 billion or as high as 94.2 billion give or take (perilously close to the 100 billion accepted figure, gasp!). Now, the +\- represents the standard deviation, and the results seem to be reliable.
Thus, in my humble opinion, the number of neurons per brain reported in the paper does not seem to be significantly different from 100 billion as a first approximation. To the author’s credit, not a big deal is made of the issue of the number of cells in the paper. Furthermore, the main subject of the paper, explicitly stated in its text (and even in the title) is not neurons, but glial cells, namely that it seems that on average, there is one glial cell per neuron. Now, to me that is the most interesting thing about the paper. Why?
When they were originally discovered, glial cells were thought to play just a structural or supporting role in the nervous system; in fact, the very word glia means “glue”. However, glial cells are much more than that; we now know that they are quite the active partners in brain physiology. They work together with neurons so there is little doubt that glial cells affect many aspects of nervous system function. I am sure that there will be more interesting discoveries down the road!
As I said, I do not hold the authors of the paper responsible for this confusion. Rather, I feel kind of annoyed when I read reports (like the one in The Guardian) that pretend to be factual without the complete, relevant information required to reach a proper series of conclusions. It may well be that the 86 billion figure is significant and it tells us something about our brain or it may as well not, but we will not know which one it is with incomplete information; that”s for sure. Already, several prominent science writers have begun to talk about the 86 billion neurons in human brains as a fact. This is quite premature and a disservice to the public and to science writing as well.
That’s why I tend to mistrust byte-size, headlines-driven science reporting. Science writers have a very serious mission.
It is not that the exact number of neurons that we may have is unimportant, not at all! This piece of information will undoubtedly help us when designing true artificial intelligence, namely by saying, how many neurons do we have? Then again, the actual number may not be as important as how all those cells communicate with each other through chemical and electrical transmission, but that’s a yet another topic for some other time.
The take-home message: Take science-related hyperbolic news with a grain of salt, and whenever possible, go straight to the source to find out what the main point was (it kind of rhymes… ).
In the meantime, I will keep using my neurons and try not to think about whether I have 80 or 100 billion of them.
Picture credit: http://1.bp.blogspot.com/
This post is based on two posts previously written for my Baldscientist blog, with some updates added and explicitly linking it to science fiction. You can find the original posts here and here.
****Already an update! I contacted Dr. Roberto Lent and he informed me that a follow-up paper with more brains is about to be published! Stay tuned!****
If you want to know more:
Azevedo FA et al., (2009) Equal numbers of neuronal and nonneuronal cells make the human brain an isometrically scaled-up primate brain. J Comp Neurol 513(5):532-41.
Fields RD (2010) The Other Brain. Simon and Schuster.
Fields RD (2004) The Other Half of the Brain. Scientific American 290:54-61.