bekolay.org

Bromine, which plays a role in a new project's chemical computer.

Slashdot just posted a BBC story about a new chemical computing project that, in some ways, mimic neurons.

The article is a bit light on details -- understandably -- but essentially their approach is to build networks of connected chemical computing cells.

The cell bodies are a mixture of liquids that undergo Belousov-Zhabotinsky chemical reactions when the concentration of bromine in the cell reaches some threshold value.

The outside of the cells is a material that the article calls a "lipid." When one layer of this lipid material comes into contact with another layer, a protein forms a passage that can transport signaling molecules between them.

I don't know much about chemical computers, but a cursory look at Wikipedia tells me that the BZ-reaction is a common element of chemical computers; the new contribution appears to be this lipid material. And it does appear to be a very interesting contribution.

While it's not explicitly stated, the tone of the BBC article is pretty clearly "hey look, we might be able to make or simulate human brains." I completely understand this urge; certainly, it's the reason it got picked up by Slashdot, and will likely find its way to other big websites. Unfortunately, it's just not true.

The BBC article touches on some of the similarities between this project's chemical computer and the human brain:

• It's made up of a network of cells.
• It shares information through chemical signals sent from one cell to another.
• The cell body processes input, and after producing an output, has a "refractory period" -- some short period of time in which input is not processed.

These similarities are minor in comparison to the sweeping differences:

• The connections between neurons are one-way: a neuron receives a number of inputs (on average about 10,000), and produces one output (though that output may be connect to many other neurons). The article does not mention whether the chemical computer's connection are one way, but it seems unlikely, for a number of reasons that I won't get into.
• A neuron communicates with other neurons by sending an action potential (a short burst of electricity), also called a "spike" due to its appearance on a graph of neuron voltage.
  
  After spiking, the neuron goes into a rest period during which it will not respond to input, called the refractory period.
  The BZ-reaction does not appear to exhibit a similar voltage graph, even though, according to the article, it does exhibit a refractory period in which input is ignored.
  
  It is possible that the particular reaction used by these researchers is more neuron-like, though I am not aware of such a chemical reaction.
• Chemical computing (as described by Wikipedia, anyway) aims to use chemicals to perform computations as we know them in computers; so, we're talking about bits and logic gates (specifically NAND gates in chemical computers). Brains, as far as we can tell, do not compute things in this way. The notion of the "bit" (binary digit, essentially a one or zero) does not even make sense in the brain, as it is a noisy system.

Of course, these differences in no way mean that the chemical computing project is not interesting and exciting. This lipid material that is able to make spontaneous connections between simple computing devices is fascinating, and while I am, again, not well versed in how chemical computers work, a huge network of these cells (perhaps 100 billion -- the number of neurons in human cerebral cortex) might have incredibly computing power.

And indeed, the quotes from one of the researchers, Klaus-Peter Zauner, tell me that it is not the goal of the project to make a human brain or human brain analogue.

The article has to go to a different chemical computing researcher, Frantisek Stepanek, to get the juicy brain quote.

And even then he only talks about "computers of similar power and complexity."

The thing that bugs me about this article is that it's not about imparting information, it's about trying to fulfill a science fiction fantasy.

The idea that this chemical computer would bring us closer to being able to make or simulate brains is a natural thing to think, but rather than futilely trying to show that it may be possible, the article should use the idea as a platform to entice the reader and then tell them something interesting about chemical computers and about how the brain works.

Information lies not only in how this new chemical computer is like a brain, it is also in how it is not like a brain. In neglecting this piece of the puzzle, this article does a disservice to the reader and to the public's scientific consciousness.