Wednesday, May 23, 2012

Biological Computers

There are some interesting similarities between computers and humans.

All computation done by computers can be thought of as manipulating numbers. Each computer program just reads and writes zeros and ones based on a set of instructions. The set of instructions (also known as the code, or the program, or the software) tells the computer when to read some digits of the input number and what to do after reading those digits (which is either read some more digits or write some digits, and go to the next instruction). Although manipulating numbers sounds like something that would be useful only for math, it turns out to be a very powerful concept since we can map almost everything to numbers (this mapping is sometimes called a digital representation). One of the key observations in computing is the fact that the programs themselves can be mapped to numbers. Any computation can be represented as a sequence of zeros and ones. The nice thing about this is that once you map a program to a number, then you can easily make copies of it just by copying the number. So it is very easy to mass produce copies of a program.

It seems like evolution also figured out that the best way to copy something is to first map that thing to a number and then copy the number. I am saying this because we can think of the DNA molecule (which is the key to the replication mechanism in all living organisms) as a number written in base 4 (each digit of the DNA number has 4 possible values. Numbers in base 4 are also called quaternary numbers, as opposed to binary or decimal numbers). The DNA number for a living organism somehow encodes all the information necessary to build that organism. Not surprisingly, for humans this is a very very large number. It is a quaternary number that contains 3 billion digits! If we were to write that number in binary, like our computers prefer to do, it would be a number that contains 6 billion digits. If you think that is very large, consider this, the number that represents Microsoft's Vista operating system is a binary number that contains 24 billion binary digits! Well, I guess at this point in our civilization we are producing programs that are more complicated than ourselves.

OK, so the first point I am trying to make is that the genetic code for living organisms is written and copied in a digital format and encodes the instructions of how to build that organism. Kind of like a program written and copied in digital format, encoding the instructions about how to execute the program.

Here is the second connection between humans and computers: The way the humans are built have some similarity to the way the computers are built. It is clear that human brain is a very special organ. Loss of no other organ would make ending someone's life acceptable, but when the brain of a person stops functioning it is somehow acceptable to turn off the life support. If you think about it, this means that we are very strongly convinced that brain is what makes us us. Brain is the center of our consciousness, it is the place where our thoughts originate and it is the organ that controls the movement of our body. Very much like the microprocessor, aka the central processing unit (CPU), of a computer. Actually, a human brain also contains that individual's memory, so it would be more like the CPU and the memory of the computer altogether.

How does the human brain communicate with the rest of the organs, how does it tell them what to do? It turns out it does this by sending electrical signals using the nervous system which consist of wires (made of cells called neurons) that establish connections between different parts of the body and the brain. This is exactly how a computer's CPU communicates with the rest of the computer. For example, the inputs from the computer keyboard and mouse are sent to the CPU as electrical signals through some wires, and the CPU sends messages to the screen by sending electrical signals through some wires. Interestingly, the first computer that was invented was not a computer that used electrical signals. The first computer was proposed by Charles Babbage and it was a mechanical device. However, he could not build it because it is very difficult to build a computing device using mechanical components. Once we discovered how to store information and send signals using electricity, it did not take too long for computing technology to flourish. Is it a coincidence that both humans and computers process and communicate information using electrical signals? Maybe, based on the physical laws of this universe, that is the best way to build a computing device.

Let me mention yet another connection between humans and computers. Noam Chomsky is a famous linguist who in 1956 developed a mathematical model for grammars that define the syntaxes of natural languages. It turns out the structure of sentences in natural languages can be described using recursive rules such as: "A sentence is a noun-phrase followed by a verb-phrase," "a verb-phrase is either a verb or a verb followed by a sentence." (Notice that a sentence is defined by referencing a sentence again, this is why these types of rules are called recursive).  The formalism Noam Chomsky developed for explaining grammatical structures of natural languages has been tremendously influential in computer science. It has been especially influential in development and specification of programming languages as one might expect. Interestingly, it has also been very influential as a model of computation. It turns out that the recursive rules that our brains use in constructing sentences can be a very effective way to describe computation. And the most general model for grammars can actually describe any computation that can be executed in the most general computing model. So, while our brains are constructing sentences, they are doing a type of computation.

I can keep going, listing more connections between humans and computers. But here is my basic point: I see humans as biological computers. In an earlier post I said that the most essential quality of a computer is programmability. I even said that anything that is programmable should be considered to be a computer. Then, in order to assert that humans are biological computers, I need to argue that humans are programmable. And they definitely are. However, the mechanism of programming humans is different than programming a computer.

Since we build the computers ourselves, we know exactly how to write a set of instructions that the CPU of a computer can execute. Because we know the exact encoding that the CPU uses (we know how it maps instructions to numbers). However, we did not build the human brain, evolution did. And, we do not yet know how the human brain encodes instructions and we do not know how to send it instructions. But we are still able to program humans. Humans are programmed by a process called "learning". I observe this process everyday watching my kids. Most of the time they learn by example, they see something and they imitate it (this is why people with kids start to use spell-speak for curse words). After enough examples and imitation they become good at whatever they are learning.

If we had a way of copying instructions of how to do something, like to play piano, from the brain of a pianist and then copy it to other peoples' brains, we would save a lot of practice time. That is what we do with computers, we copy programs from one computer to another. And immediately the computer that gets the copy of the program becomes as good as the previous computer in executing that task. But, someone has to write the program first. This is where humans have superiority. You see, the human brain writes the program of how to do something itself during the process of learning. It somehow figures out the instructions for how to play piano, how to speak, how to write, how to play tennis, etc. by just observing and imitating. Interestingly, learning as a way of programming has become a very influential topic in computing in recent years. The general area is called "machine learning". The basic idea is to make computers learn how to do something by observing lots of examples. This is how Google Translate works for example, it uses a machine learning algorithm that figures out how to translate a sentence from one language to another language by looking at existing translation examples on the web.

The basic idea that computers in some ways can imitate humans has a long history. There is an area of study in computer science called "artificial intelligence" which has been investigating how to make computers "intelligent" like humans since 1950s. However, artificial intelligence does not sound very futuristic anymore since computers were able to beat  the best human players in chess and Jeopardy! The problem with the term "artificial intelligence" is that it seems very hard to define what intelligence is. We tend to call people who are good at analytical reasoning and math intelligent. Computers are much better than humans on some analytical and mathematical tasks, so one can claim that computers are very very intelligent. But they do lack something. They do not have consciousness. Although consciousness is yet another thing that is hard to define clearly, I personally consider it to be the most significant difference between humans and computers. Humans are aware of their existence whereas computers are not. Is this something humans are born with? Or is it something they learn? Is it something that can be programmed to a computer? Or is it something computers can learn in some way, maybe using machine learning? Time will tell. But, if a computer gains consciousness someday somehow, I bet turning its power off will become a tough decision.


  1. Viewing humans as programmable through learning is a very interesting way of interpreting how the human brain functions. I can see that the general theme throughout this seminar and class is that everything can be defined in the terms of a computer, but this also facilitates how we understand computers. By using these several analogies, it becomes much easier to interpret how computers work.

    Also, I'm very sorry for being unable to attend the last seminar section. Sorry!

  2. Viewing the human brain as a programmable computer through the process of learning is an intriguing take on the comparison of the brain and computers. At one point you mention how it would save us so much time to copy things learned in one brain and paste them into others... With all of the similarities between the brain and computers (transferring data and signals via electricity), it seems that that might not be such a outlandish idea.

  3. Biological computer is a very fascinating topic. I wonder if using human brain to create computer is actually possible? In movies, some of them talk about computer built on the basis of human brain which were used to control all sorts of dynamics, is that possible?

  4. Sorry to comment this article late. Somehow I didn't see this article under the may section. According to your article, you said that "computers were able to beat the best human players in chess and Jeopardy.". I was very surprise that how can a human made artificial intelligence had the capability of exceeding the makers. I heard that computer has some possibility of making error. In this case, is that mean human error is more critical than computer's?

    -Zhuocheng Yu