Robots have come a long way in the past 85 years or so. The first “Robot” appeared in a Capek play in 1920 as an artificial worker. The word was derived from the Czech “robota,” meaning labor.
In the 1940s and '50s Asimov created the intelligent robot R. Daneel Olivaw, indistinguishable in looks and behavior from humans. Asimov did not specifically equip his robots with computer brains since computers were still thought of as fancy calculators and Artificial Intelligence had not yet been defined. That happened during the 60s, and the computer HAL 9000 in Clarke/Kubrick's 2001 Space Odyssey became a famous example of a computer with an intelligent brain, seemingly capable of using its own free will.
To this day artificial intelligence remains an unresolved problem. The question remains to what extent a computer will ever be able to simulate the workings of a biological brain. Although they are sometimes called electronic brains, computers have nothing in common with biological brains. Electronic computers are binary digital, in that the basic elements can only be "on" or "off." A biological brain is analog and its basic elements (synapses) can assume a continuous range of values. The materials of construction are totally different. The complexity of the brain is orders of magnitude greater than that of even the most advanced computer. And the most important difference lies in the way information is processed.
Electronic computers are literal-minded idiots. They need a program that tells them step-by-step how to do a task, and they'll follow their instructions in sequential order. With a given input there is just one possible result (although not always what's expected!). A robot with such a brain is totally predictable – no whimsy, no learning from experience. It never gets bored either.
On the other hand, the 100 billion or so connections in the human brain, called synapses, can assume many states, from off to fully on. They are also very highly interconnected, with some of them connected to as many as 10,000 others, and the state of each synapse depends on the states of the synapses it is connected to. In other words, the brain operates in an extremely parallel fashion. Every thought and every observation affects millions of the synapses in a process we know as learning from experience.
The structure of the brain makes it a relative slowpoke in performing the straightforward serial operations that electronic computers excel in. But because it learns from experience the brain excels in performing very complex tasks that are almost impossible to achieve with electronic computers. I can locate my wife in the back of a crowded store from a glimpse of the back of her head. I recognize the sender of a letter from his handwriting. I know the difference between happy and sad, love and hate, loyalty and treachery. Try that on your laptop. They are tasks I know how to do without knowing how to provide step-by-step computer instructions.
The mode of operation of the brain has been emulated with electronic computers implementing so-called neural networks. The way synapses interact and modify each other's states has been simulated in software modules, called neurons, combined in highly parallel networks. Even with a very limited number of neurons these systems achieve startling results in performing tasks that are almost impossible to achieve with traditional programming.
Neural networks are used in applications such as vehicle control, handwriting recognition, medical diagnosis, chess games, facial identification, and email spam filtering, amongst many others. Significantly, they are not programmed in the traditional way but are trained by example and experience, trial and error, similar to the way living beings learn. Values of the neurons resulting in correct answers are increased and values leading to incorrect answers are decreased. It is frustrating that there is no way of telling how and why the neural networks perform as well as they do, even when the number of neurons is fairly small.
It is not inconceivable that someday someone will implement the hardware equivalent of biological synapses connected in a 3-dimensional network to simulate a biological brain. What all could we expect of a robot with such a brain? Given visual, auditory, and tactile sensors, it would almost certainly recognize its operators. But would it appreciate a good joke? Without the powerful stimulants of sex, greed, and ambition, what would motivate it, if anything? We might expect it to be superior at logical reasoning, but could it have artistic creativity? A possible Einstein, but never a Beethoven?
Intriguingly, would it have a will of its own to distinguish it from the robots with a serial electronic brain? It is difficult to see how this could be. Free will assumes that there is a choice between alternatives, a fork in the road. Who or what is going to make the decision which road to take? If the brain's hardware determines which road is preferable, the outcome is predetermined and there is no free will involved. If the choice is made randomly the outcome is not predetermined, but we can't call that the action of a “conscious” free will. Scientists, philosophers and theologians have not yet agreed on what this consciousness is, or if it is a figment of our imagination.
Humans feel certain that they can do things “out of their own free will,” even in the face of the compulsive behavior of addicts, repeat criminals and persistent nail biters. We make some important decisions over the years that affect us the rest of our lives. Choices of schooling, marriage, career, emigration for some of us. What is it in our brains that set us on the chosen courses? If our brains did not determine the outcome based on the states of its synapses, what was it that chose which road to follow? Maybe after all, we too “ain't nothing but robots?”
Think about this next time you are tossing and turning in the dark. You'll be asleep in less than no time! (Inspired by articles about the brain in TIME magazine of 1/29/07)
This article has been provided to APCUG by the author solely for publication by APCUG member groups. All other uses require the permission of the author (see e-mail address above).
For more information on the Tulsa Computer Society click here