We must set the limits within which Artificial Intelligence can operate. If we fail to do so and ignore this duty, we shall be accomplices to its consequences. That is the view of the world’s leading experts, companies and institutions. We are talking about people such as Stephen Hawking or Elon Musk, and companies such as Amazon, Facebook, IBM or Microsoft. The sooner we have this debate, the better.
Let me be clear: I am passionate about technology and a firm advocate of progress in all these fields. The reason is very simple. Thanks to how far we have already come, and how far we still have to go, it is within our power to create a better, healthier, fairer and more caring world, one that lays the foundations for what is undoubtedly our next frontier: not the Earth, but the Universe.
That said, the impact of Artificial Intelligence goes far beyond that of a mere tool in our hands. We are creating the origin of something that has never existed before. It is “something” that learns by itself and may come to make its own decisions without those decisions necessarily being controlled by humans. It may even carry out actions that were never “programmed” by its creators. In other words, we are venturing ever further into the creation of “something” capable of a will of its own.
There is much talk of Artificial Intelligence, but what is it really? From a technical point of view, Artificial Intelligence is defined as a computer’s capacity to perform operations consistent with human learning and decision-making. From a scientific point of view, the human being is a prodigious thing. I have always been amazed by an aunt of mine’s innate ability to look at a newborn baby and declare with absolute certainty that the little nose is the mother’s, the eyes the grandfather’s, and the little ears those of some second cousin who had exactly the same ones. Well, behind this apparently so “scientific” observation lie processes of learning, comparative calculation, memory and decision-making that are utterly prodigious. Only in recent years, with heavy investment and highly advanced computing, have we begun to have reliable computerised facial-recognition systems outside the usual science-fiction films and crime-solving TV series. What humans (mothers-in-law included) can do with the greatest naturalness and no apparent effort conceals processes that are the basis of what we know as Artificial Intelligence.
Remember: the basic process is learning, and decision-making based on that learning. This is where the computing power and memory of computers (in the broad sense) come into play. There is no comparison with human capacity. It is vastly greater, and the gap between what a computer and a human being can do is set to widen exponentially. Quantum computers are already here, and they will represent a tremendous qualitative leap, making today’s powerful computers look utterly ridiculous. What does this mean? That a computer’s capacity to learn is, and will increasingly be, far superior to that of a human being. That means the capacity not only to memorise data but to process it and find relationships within it: what technicians call pattern recognition. You already know what my aunt would say about that.
Once we have found patterns, the next step is to analyse a new situation, see which pattern it follows or most resembles, and thus be in a position to predict how that situation will evolve. To return to my dear aunt: some time ago, poor thing, she broke her hip. Ever since, whenever it aches, she notices that it tends to rain. So if she wakes up with that feeling and has to go out, she takes an umbrella, even if the sun is shining. In other words, were she a computer running Artificial Intelligence systems, she would have observed this pattern and, on detecting the pain, would conclude that rain was highly likely that day. Now we come to the crucial point, the one that can truly change the world: decision-making based on that learning and on the discovery of patterns we have just described. My aunt’s case is fairly harmless and goes no further than carrying an umbrella. In the case of a computer without limits, it could be very serious. Cinema offers many examples, which fortunately have remained on celluloid. Who does not remember HAL in 2001: A Space Odyssey, which decided that its “life” was more important than that of the humans? Or the Skynet system in Terminator, which concluded that humans were a danger and that it had to take control and eliminate them? Among more recent offerings I recommend Ex Machina, which explores the creation of a humanoid “robot” (Ava) with Artificial Intelligence and its evolution until it becomes aware of its own existence. I encourage you to see the consequences in the film, but I would not want to give you a “spoiler” here.
IBM is well advanced in this area and has a multi-million-dollar project that it has been developing for several years. Its origins lie in the famous versions of Deep Blue that were the first to defeat chess grandmasters. It is called Watson, and as far back as 2011 it managed to win a TV quiz show called “Jeopardy!”. Last year Microsoft had to shut down a bot (a social-media robot) called @tayandyou, designed with Artificial Intelligence to hold automated conversations on Twitter… because it turned racist!
All this remains essentially harmless, entertaining and in some cases even amusing, but the foundations are there. The same process used to learn and make decisions on social networks or in a TV game show can also be used to decide whether a person is a potential terrorist and should be eliminated, to bomb a conflict zone, to steal your identity, to tamper with a nuclear power plant, to alter a surgical procedure or to change the substances added to drinking water.
“Terminator Genisys” (Credit: Paramount Pictures)
As long ago as the last century, the brilliant Isaac Asimov set out his famous Three Laws of Robotics. Perhaps we need to return to them, no longer from a literary point of view but from a legal one. Otherwise, we already know the consequences: “I’ll be back!”
Francisco Canós
Article published on 29 July 2017 in:
