If we listen to Elon Musk, the visionary tycoon who founded PayPal , the electric vehicle and battery company Tesla or the private space corporation SpaceX , humanity faces a new and formidable threat: artificial intelligence (AI). “It’s like those stories where someone summons the devil. There is always a guy with a pentacle and holy water convinced that this way he will be able to control him , and of course, it doesn’t work, “he says.
Your concern has a lot to do with money. The heavyweights of the technology sector are betting heavily in this regard. Google, for example, acquired the company DeepMind, specializing in the development of neural networks in which Musk had already invested. The search giant works on a computer system capable of distinguishing in a video a human face from that of a dog, people skating or sleeping, a cat … And all by itself and without anyone having put labels on the file previously. The investment in artificial intelligence is such that today we can enjoy digital assistants that facilitate many day-to-day tasks such as Echo from Amazon.
The idea is that he learns, so to speak, after ‘feeding himself with millions of recordings’. IBM, for its part, fine-tunes its Watson supercomputer, which in 2011 defeated the human champions of the American question-and-answer contest ‘ Jeopardy!’. Its intention is to improve the cognitive functions of the device and to test its abilities to perform medical diagnoses, personality analysis and translations in real time. Facebook engineers are not far behind and have devised an algorithm that allows you to successfully recognize a face 97% of the time, even if it has been poorly captured.
Musk says things are going too fast, and that is why AI is a technology that can be as dangerous as nuclear briefcases. In the chorus of doomsayers of the artificial apocalypse, the voice of the British philosopher Nick Bostrom , from the University of Oxford, stands out, comparing our destiny with that of horses, when they were replaced by automobiles and tractors. In 1915, there were about 26 million of these equines in the United States. In the 1950s, there were only two million left. The horses were euthanized to be sold as dog food. For Bostrom, AI poses an existential risk to humanity comparable to the impact of a large asteroid or the nuclear holocaust . All of this, of course, provided we can build thinking computers. But what exactly does this mean?
The future predicted by fiction
In reality, the concept of artificial intelligence is not as recent as it seems. Since the time of Alan Turing – who is considered the father of it – and the construction of his Bombe device, which made it possible to decipher the codes of the German Enigma machine, more than seventy years have passed. At one point in the film ‘ The Imitation Game’ (Morten Tyldum, 2014), in which Benedict Cumberbatch plays the famous mathematician, a detective asks him: ‘Can machines one day think like humans?’ . To which he replies: ‘Most people think not.’
The problem is that you are asking a stupid question. Of course, machines cannot think like people. They are different, and they think differently. The question is: ‘Just because something is different, does that mean you can’t think?’ The detective then asks him about the title of his article, ‘The Imitation Game’ . “It is a game, a test to determine if someone is a human being or a machine,” says Turing. ‘There is a general theme. A judge asks, and from the answers, decides if he talks to a person or a machine . ‘ The scene may be made up, but its content is real. The test exists.
AI is all the rage thanks to literature and movies. But what is the actual degree of progress? Years ago, I was at the Pittsburgh Institute of Robotics , in the USA, one of the temples of this discipline. At that time, he was part of a TVE team that collected the latest techno-scientific advances in a dissemination series called 2.Mil . I have to admit it: I got a big disappointment because of the image of robotics that science fiction has instilled in us.
The gadgets they had there were little more than pots in the hands of engineers in jeans, and looked like they had come out of a geek garage. They broke down at the slightest occasion. They told me about Florence , a robot nurse who was going to revolutionize geriatrics. Actually, it was a kind of barrel with a head to which silicone eyes and lips had been glued to draw smiles.
Florence had a built-in television camera and monitor. Batteries were draining fast. And, of course, he did not understand what we were saying to him. Everything he spoke had to be programmed in advance, so an engineer worked hard to get him to come out into the hall and give us a welcome message.
He had read a lot about what they did in Pittsburgh, especially from Xavier , a robot who knew where he was going, a complete revolution. But it was nothing more than another barrel with wheels that moved through the corridors of the institute thanks to a map that he had in his memory. Before some stairs, he stopped so as not to kill himself. Apparently he was breaking into the premises to tell dirty jokes. That morning I saw Xavier being dragged away, an image I will never forget. I was in the catacombs of robotics! I went to the office of Hans Moravec, one of the most famous visionaries, but everything he said was hard to believe.
Moravec was convinced that in fifty years the androids would displace the humans. For more than an hour he talked non-stop about the evolution of these devices and their growing intelligence, thanks to the advancement of microprocessors and their ability to handle more and more information. It was a captivating talk. The evolution of the machines was going to be unstoppable. “The time has come for us to leave,” concluded the Austrian-born scientist.
Moravec left the institute to found a company for industrial robots with 3D vision. Before, he had shown me on his computer an image perceived by one where chairs and tables that had a pixelated look were seen. And how could the machine know what was what? In that summer of 1999, Moravec said that he was fascinated by a new Internet search engine, the smartest and best designed . It was the first time I heard of Google.
In 2014, Google bought an AI company from Musk and has developed the first autonomous car, which has already traveled a couple of million kilometers without a driver, and the system that differentiates cats from people on YouTube. The world is literally invaded by inconceivable amounts of information circulating on the network and computing capacity is constantly increasing. But do we really have reason to fear that a machine will one day think like us?
Ramón López de Mantarás, director of the Artificial Intelligence Research Institute of the CSIC, is one of the most recognized Spanish experts in robotics and IA ‘No’, responds to VERY. And he adds: ‘I do not know what will happen in hundreds of years, but all this topic that is talked about, the singularity, the transcendence, that there will be machines with consciousness and improved qualities with respect to human intelligence in a matter of thirty or forty years Has no sense. .. I have never seen a scientific argument to support it. ‘
López de Mantarás’ point of view fits with what I felt so many years ago, on my frustrating visit to Pittsburgh. The world has changed a lot since then, but the truth is that the machines that will end up being aware of themselves to unleash the catastrophe are not yet in sight , as is the case in the Terminator movies.
He disagrees with gurus like the futurologist Raymond Kurzweil , who today works in the Engineering division of Google. Like Moravec, he is convinced that during this century robots will be able to pass the aforementioned Turing test, even before 2029. Well, it would not be the first time that one of his predictions has been fulfilled.
At the end of the 1980s, Kurzweil claimed that around 1998 a computer would beat a world chess champion: it happened in 1996, when Gari Kasparov lost a game against the IBM Deep Blue software. In those years, he also imagined that the Internet, by then a network relegated to academic institutions, would spread throughout the world. He now claims that in 2045 computers will be far more powerful than all the human brains on Earth put together. When he speaks, the media falls in love with this artificial world. “He is a very media person,” concedes López de Mantarás. “But behind there is nothing scientific.”
Marvin Minsky, co-founder of the Artificial Intelligence Laboratory of the Massachusetts Institute of Technology (MIT) , awarded by the BBVA Foundation, does believe that machines as intelligent as humans will be developed. “However, how long this takes will depend on whether you work on the right problems and money,” says Minsky. “He is a terrible enfant ,” says López de Mantarás, who was at the congress that MIT held in his honor.
“Minsky thinks that the great advances in this field were made between the 60s and the 80s, and then all ideas regarding AI in its general sense were abandoned.” Thus, what would have remained in the current panorama is specialization, machines that are extraordinary playing chess, but that do not know anything about checkers or Parcheesi . ‘Specialized AI is good business, and I’m in favor of it. It would be more. It is what artificial intelligence really is today, ‘says López de Mantarás.
Generalist AI research is disappearing. In this world awash in data, this technology is completely different. Google’s autonomous car or IBM’s Watson supercomputer analyze terabytes of information to make the right decisions. However, they do not know how to explain how they got there. In other words, when the system spits out its response, it is unable to answer the question, why is that? “We have renounced the why and we have stayed with the what,” laments Pérez de Mantarás.
For example, a few years ago, an expert system that deduced that the patient had pneumonia could justify that diagnosis from the patient’s history and the cultures performed , but now a software as complex as Watson cannot do it. It simply draws a conclusion from the overwhelming amount of data it handles, but does not offer a reason. “This creates acceptance problems for the user,” says Pérez de Mantarás.
Let’s look at the movie ‘Yo, robot’ (Alex Proyas, 2004): the streets are full of humanoids who carry the purchase, serve drinks, distribute hot dogs … If we forget for a moment the fantastic agility they demonstrate and the general intelligence that they treasure, what would we have left? It is clear: an army of specialists.
Drones, a manifestation of this trend
The US Air Force maintains more than 8,000 of these devices to combat terrorism, according to the Brookings Institution. In their operations, they have already killed more than 2,500 people. Business models, for their part, film and investigate anything. For example, equipped with infrared sensors, some can detect which plants are sick or suffering from parasite attacks.
Thus, it is possible to devise a fumigation plan à la carte. Others help control poaching and provide clues to biologists studying bird flights and trajectories. They are of all kinds. The largest of all, the Israeli-made Eitan, has a wingspan of 26 m, almost like a Boeing 737. In contrast, the diminutive 16 cm Nano Hummingbird, developed with the support of the Research Projects Agency Defense Advanced of the United States (DARPA), could pass for a hummingbird. The robots have literally taken off to their freedom, although, yes, under human control.
Among the robotic fauna that remains on the ground, we can find from the Roomba, a small semi-autonomous vacuum cleaner in the shape of a disk designed by iRobot that has become a commercial success, to the PackBot -of the same firm-, a military rover equipped with a robotic arm capable of manipulating bombs to inspecting places contaminated by radioactivity . The TUG, for its part, the work of Aethon, looks like a kind of table with wheels and sensors. Thanks to the latter, he moves seamlessly through the corridors of some American hospitals to carry medicines and other supplies.
