By Harry Keller
Editor, Science Education
You may have noticed recent news about Stephen Hawking predicting the demise of the human race due to the rise of artificial intelligence (AI)1. Others of genius rank, such as Elon Musk and Ray Kurzweil2, have also made this prediction. With “The Theory of Everything” (biopic about Prof. Hawking) in theaters right now, this prediction is resonating across the English-speaking world.
Before digging your shelters or heading for the hills, you should ask, “What is artificial intelligence?” A bit of history may help put this entire subject into perspective. The term was coined in 1955 by John McCarthy, who called it “the science and engineering of making intelligent machines.”3 “When?” you ask. “That’s nearly 60 years ago! Before I was born!” (I actually was born well before then, but statistically you probably weren’t.)
Funny how AI has not taken over the world in the 60 years it’s had so far. Why the sudden worry? Computers are becoming more powerful and less expensive. Computer memory is dirt cheap, speaking historically. This trend of more computer power and more memory shows no signs of abating soon. Could it eventually reach the tipping point where machines are sentient and self-reproducing? Would they then remove the “scourge” of humans from the Earth’s surface? Might the end be less dramatic in that they would render people superfluous? Imagine a world in which all work, including creative work, is done by machines. Who needs Beethoven when you have the Ultra-Composer Mark IV?
This entire discussion circles around to defining machine intelligence and estimating exactly how smart machines might become. Right off the bat, understand that intelligence, as we commonly understand it, has not been seen in machines yet. No one truly knows if it ever will be. To comprehend why, you must have a feeling for the nature of computers and computing.
These days, we are seeing a movement in the direction of teaching coding in schools, even in kindergarten! While I consider this effort to be foolish, I do support teaching the nature of computers and computing throughout our K-12 school systems. A computer is, in essence, an idiot savant. That is, computers have an IQ of around zero but have an inhuman ability to remember and recall information precisely as well as the capability of performing calculations at speeds that are literally inhuman. As with John Henry and the steam-powered hammer4, we don’t stand a chance at outperforming a computer in calculation speed. (Legend has it that Henry beat the drill but died soon afterward, a pyrrhic victory indeed.)
More recently, Deep Blue beat world chess champion Garry Kasparov5. Is Deep Blue intelligent? Not a bit. It was programmed to search out all possible moves and following moves for from eight to 20 moves deep as well as having all possible book openings and end games memorized. This was not intelligence; it was simply brute strength applied to what used to be a subtle game of strategy.
Fast computation does not make one intelligent. The other part of the nature of computers and computing is that you have to tell a computer exactly what to do. Without instructions, a computer is just an expensive door stop. So far, only humans have been able to master the complex task of engineering software to make computers operate, and they haven’t been too great at that if you consider all of the bugs (and security holes) found in software at any point in the history of computing.
Creating a self-aware machine capable of independent thought is a project that makes every project ever done by the human race, from the great pyramids to the Apollo program, seem minuscule in comparison. No one even really knows where to begin. They have some clues, it’s true, but no one can lay out a project plan for creating an intelligent machine today. All of this talk about AI taking over the world is just so much science fiction until someone can point to a real path to getting there.
Assume that such a path were laid out and blessed by those who are capable of evaluating such things. It would certainly be a decades-long project. Even were it not, it would still take so long and so many people that it would be filled with flaws. Managing this project would be a nightmare. Who would enable and pay for this project? Why? We have no pressing requirement for machines that can match or exceed human intelligence. Instead, we are seeking to program computers to do what we do not do as well as we’d like. Designing airplanes, to take one example, requires enormous numbers of calculations of stress and strain that no person could possibly do. The computers that aid our aeronautical engineers in this task are hardly intelligent by any definition.
Ask yourself if you believe that a computer could write this article, given the topic. If you answered yes, then consider exactly how you would program it to do that. How would it select the sequence of ideas to present? How would it write its sentences? How would it choose the words? You have to break each action or decision down to very simple steps, steps that a four-year old could do and then even simpler. Don’t feel bad if you don’t have the answer. No one does.
Kurzweil and others notwithstanding, we have little concern for the future with regard to AI. One of the things that bothers Prof. Hawking is that machines can evolve much more rapidly than can humans. Ignore for the moment that any computer has been programmed by fallible humans and must have errors in it that will likely be magnified if it tries to expand its capabilities or even to remove the errors. The current pace of research in genetic engineering holds the real potential for the engineered evolution of the human species. Other than eliminating some diseases, we don’t really know what to do with this newly emerging power. Should we engineer more intelligence, greater running speed, more stamina, or increased musical ability? We might eliminate the “tone deaf” gene, if there is such a thing.
While many unanswered questions remain, we can evolve faster than nature allows. We also can put limits on what machines can do because we are the ones who create the instructions that tell them what to do. We control the power switch too. Human beings are in charge of machines. We are not in charge of what others do with them.
The problem of abuse of computers by others is real and must be addressed. Your personal information is being hacked. The “Internet of Things”6 could end up being the most intrusive spy network every. And so it goes.
Let’s not worry about way-out-there science fiction scenarios about AI taking over the world as in “Terminator.” Instead, let’s put our human intelligence to work preserving our privacy and security against those who would use machines against us. Let’s also train our students to understand about computers (actually coding them being unnecessary) and to develop strong critical and creative thinking skills so that others cannot fool them into ceding their independence to those who control the biggest machines such as search engine networks and media conglomerates.
1Rory Cellan-Jones, “Stephen Hawking Warns Artificial Intelligence Could End Mankind,” BBC News, 2 Dec. 2014.
2 Tanya Lewis, “Stephen Hawking: Artificial Intelligence Could End Human Race,” LiveScience, 2 Dec. 2014.
3 Wikipedia, accessed 5 Dec. 2014.
4 Wikipedia, accessed 5 Dec. 2014.
5 Wikipedia, accessed 5 Dec. 2014.
6 Wikipedia, accessed 5 Dec. 2014.