The history of technology is full of predictions that sound ridiculous now. One of the most famous examples is attributed to Bill Gates, who declared in 1981 that "640 kilobytes should be enough for anyone." AI predictions are no different in this regard.
The first researchers of AI (artificial intelligence) believed that we would have a robot that would walk, talk and think like a human after only a few decades. Of course, despite some impressive advances in machine learning, AI still has a long way to go. According to a principle known as the Moravec paradox, we can teach machines to solve complex problems, but at the same time they cannot cope with the simplest problems.
Come on Siri, think like a kid
In 1957, economist and computer science pioneer Herbert Simon said: “It is not my goal to surprise or shock you, but I can sum it up by saying that there are machines in the world now that can think, learn, and create. In addition, their ability to perform these actions will grow rapidly until (for the foreseeable future) the range of problems that machines can handle is comparable to the range of problems where the human mind has so far been needed.”
Simon died in 2001, and his "visible future", in which machines can think like humans, is still a long way off. Sure, artificial intelligence has proven itself well for performing specific tasks like classifying distant galaxies or mimicking celebrity voices or creating art, but simple thinking - a concept known as general artificial intelligence - seems to baffle the most advanced machine learning systems. Just think, even walking on two legs is a challenge for machines. They may be able to defeat the great chess champion, but they will not be able to get ahead of the little one and take the right toy from the shelf.
This is not a new problem. In the 1980s, computer scientist Hans Moravec presented precisely this problem, now called the "Moravec paradox," and explained why this is exactly what we should expect from machines that are not subject to natural selection. “Coded in the large, highly developed sensory and motor parts of the human brain are billions of years of experience about the nature of the world and how to survive in it,” he wrote in his 1988 book Children of the Mind.
Promotional video:
That is, what seems simple to people has been improved for millennia in the process of evolution. What people find the most difficult is difficult only because it is new to them - we have been thinking about chess strategy for a little over a thousand years, but we have been learning to interact with the environment since our ancestors were still single-celled organisms. Evolutionary skills do not require conscious thinking, and when you don’t need to think about something, it’s harder to figure out how to teach a machine to do it.
Get to know machines by getting to know yourself
So how do you teach a machine to really think? Moravec believes that machines lack evolution. However, the situation is improving day by day.
Engineers are teaching artificial intelligence algorithms, such as teaching robots to play video games. But before we can teach machines to think like humans, we ourselves need to better understand how humans think, understanding the limitations of machine learning can help answer questions about how our minds actually work. It is also possible that the paradox is that AI will never be truly independent and will always rely on human help. But in any case, we should all appreciate the supercomputers running inside our skulls. They make the world's most difficult tasks look easy.
Svetlana Bodrik