While the great majority of people believe that humans have free will, virtually no one thinks that computers do. But consider the following thought experiment: Assume that technology advanced to the point that a functioning computer could be built in a lab that consisted not of silicon-based semiconductor chips but of billions of flesh-and-blood neurons, molecule for molecule and cell for cell identical to a human brain. Would such a computer have free will?
Given that the quality and capabilities of artificial intelligence are converging with those of human intelligence, are there any fundamental reasons why computers could not achieve free will? Following are four arguments for why computers do not have free will. Perhaps analyzing those arguments will shed light on the more fundamental question of whether free will in humans actually exists or is illusory.
(a) Free will can only arise at an extremely high level of cognitive sophistication. Computers lack that level of sophistication.
In 1950 the British mathematician Alan Turing posed a test, which came to be known as the Turing Test, that he believed would, when passed, demonstrate that computers are intelligent, at least in the way that humans are intelligent. Assume, he proposed, that you were asked to interrogate two individuals, one of which was a human and the other a computer, not knowing which was which. If your success rate in spotting the computer was below a certain level (Turing suggested 70%), then the computer could be said to be intelligent. To my knowledge we have not yet succeeded in creating a computer that can reliably pass the Turing Test.
But I do not believe that the Turing Test is a fair intelligence test for computers. While the test might serve for determining if a computer is intelligent in a broad sense, that does not mean that computers are not intelligent in a narrower sense. Every handheld calculator can do simple math, and many can do complex math, well beyond the skills of most humans. Is that not being intelligent?
And then there was Deep Blue, a computer program that in 1997 defeated the world’s then best human chess player, Gary Kasparov. During the match, Deep Blue was considering different chess positions at the rate of 100 million to 200 million—per second. In the 20-some years since then, computer chess programs have only gotten better. Competitive chess players use such programs (known as chess engines), not so much as opponents but as tools for improving their game against other human opponents. They could not realistically hope to defeat the better chess engines in competition. Clearly such computers are exhibiting a great deal of intelligence, just in a specialized area.
But computers are demonstrating substantial intelligence in a broader sense as well. Watson is a program that was developed to play the TV game Jeopardy! Anyone who has watched Jeopardy! understands that it requires an encyclopedic knowledge in a broad range of trivia subjects as well as an ability to comprehend idiomatic expressions, puns, and other plays on words. In 2011 Watson played an exhibition match against two of the best former champions of the game. Watson had to compete in real time alongside the two humans and did not have access to the internet—all of Watson’s knowledge had to be stored in its own memory banks. The only concession was that Watson received its clues in writing over the same time frame as the clues were being read orally to the human contestants. Watson demolished its human competition. At the end of the match Watson’s score was more than triple that of each of its opponents. Obviously, Watson was demonstrating a very high—and broad—level of cognitive sophistication. That was eight years ago. Now voice recognition computer programs that answer questions and follow commands have become everyday household devices.
(b) Computers do not make choices. Instead, their output is dictated by the input and the constraints of the program under which they operate. They are essentially robots.
If you ask a handheld computer to multiply 6 x 23, it will invariably generate the result 138. In performing the calculation the computer does not exercise any choice in the matter. While this is true for simple programs and simple tasks, it is not the case with more complex programs such as chess engines. Such programs do not mechanically dictate every move for every position in a game. There are just too many possible positions to consider and choices to deal with. Instead, such programs establish certain strategies but leave to the computer the decision of how to pursue those strategies. And in fact the choices the computer makes are not predictable or determinable in advance. Instead, they are decided by a randomization process built into the overall program, as well as the particular strategies that the computer chooses to pursue. Thus, for example, chess engines do not always make the same opening move, even though the board position at the start of every game is identical. In short, sophisticated computer programs are not strictly robotic; they make choices, just as humans do. But does that mean that either computers or humans exercise free will?
(c) Free will exists as an emergent property stemming from the physical composition of the human brain. Computers are made of semiconductor chips, not biological neuronal networks.
The human brain is the result of over 500 million years of evolutionary history, a history that has resulted in powerful advantages and disadvantages. With its 85 billion neurons and perhaps as many as one trillion synaptic connections, the human brain has enormous versatility that has enabled us to survive and thrive in a wide variety of environments. That versatility has also provided us the ability to acquire knowledge in areas, such as philosophy and astronomy, that are not immediately critical to basic survival.
On the other hand, that same evolutionary history has created a cognitive system that, because it is made of flesh and blood, is vulnerable to malfunction and system failure and can be woefully inefficient, at least as compared to computers based on semiconductor chips. Witness the computational speed of even the most basic of computers compared to human abilities. Recall, for example, the power demonstrated by Deep Blue in chess and Watson in Jeopardy!
Having said that, I can think of no reason why a human brain would have any quality that, by its physical nature, would qualify it to possess free will. Although it is built of flesh and blood, the brain at its most basic level is comprised of the same building blocks—molecules and atoms—as everything else, the same building blocks that are governed by the rules that operate in the rest of the universe. These rules simply do not contemplate that some other force (viz., free will) operates solely in the nervous system to override the rules that otherwise govern the remainder of the universe. And there is no objective evidence to support such a force.
(d) God specially endowed humans with free will. The rest of the world, including computers, does not have it.
This notion that God invested humans with free will is essentially a religious argument. It is not based on scientific evidence. It is not, therefore, subject to the same logical discussion that applies to questions of a secular or scientific nature: One can always invoke the assertion that God did it—end of argument.
However, no objective evidence exists to support the idea that humans have free will. The most that we can say is that brain activity is incredibly complex and that we have a powerful subjective feeling that we are in control of the choices we make (whatever is meant by the expression “in control”).
As pointed out above, in all of the observational research done on the human nervous system, there is no evidence to support the existence of some sort of force that disrupts the electro-chemical processes that occur in the human nervous system. The laws of nature operate in the same fashion within the human nervous system as they do everywhere outside of the nervous system. There is no room for a “free will force.”
© 2019 John M. Phillips
Wow very interesting to read
ReplyDeleteThanks, Wanda.
Delete