In this blog post, we look at the reasons why the development of artificial intelligence is slower than expected, not as a technical limitation but as a problem with the approach, and seek the answer in the nature of human intelligence.
In his 1988 book Children’s Minds, world-renowned robot theorist Hans Moravec predicted that robots would become so intelligent that a new generation would emerge every ten years in the 21st century, with the fourth generation reaching human levels by 2040. However, his prediction was off the mark. He expected that the first generation of robots developed by 2010 would have the intelligence of a lizard and be able to clean bathrooms, mow lawns, and cook delicious meals in the home. However, in 2024, well past 2010, the only robots we see in our daily lives are robot vacuum cleaners. Depending on your point of view, even robot vacuum cleaners are merely machines that move according to sensors, and it is difficult to consider them intelligent robots. IT technologies, such as the rapid development of the Internet and smartphones, are advancing at a pace that meets or exceeds people’s expectations, but the development of artificial intelligence technology is relatively slow. Is this simply because artificial intelligence technology is more difficult than other technologies? Of course, the answer to this question may be “yes,” but I would like to point out that the current approach to AI technology is wrong.
Many proponents of AI view the brain as a kind of well-designed computer and are researching AI technology as an extension of computer technology. Rather than focusing on how the actual human brain works, they are focusing only on behaviors that appear to be the result of human or animal intelligence. They believe that artificial intelligence can be realized when computers have sufficient memory and processing power. However, this is a very inefficient approach. For example, consider the simple act of picking up an object such as a wallet. Humans can pick up a wallet without thinking about how to do it. However, current robot technology requires calculating the shape and position of the wallet, the shape and angle of the robot’s hand, and the force required to pick it up. Catching a flying object is even more complicated. The trajectory of the object must be tracked at every moment, and the above process must be repeated at the moment it arrives. This process is much more complicated than it seems and cannot be easily achieved with current technology. Of course, considering the rapid advancement of computer speed, robots like the one above may appear in the near future. However, is it really right for machines to perform tasks that humans can do unconsciously based on previous experience without any calculation, by performing a huge amount of calculations every moment?
How is the human brain able to effortlessly perform tasks that are extremely difficult for machines to process? The answer to this question remains a mystery that scientists must uncover in the future, but one thing is certain: it is not because the brain is faster than a computer. Neurons can be considered the brain’s equivalent of transistors, which are the basic units of binary computation in computers. Neurons receive input from synapses (the junctions between neurons), combine them, and decide whether to send the output to other neurons. Neurons usually complete this task within about 5 milliseconds and then return to their original state. In other words, they perform this task 200 times per second. This may seem fast, but modern silicon-based computers can perform such operations 1 billion times per second. Machines are already fast enough compared to the human brain. However, they still cannot match the human brain in terms of intelligence. This fact tells us that increasing speed is not the answer to achieving human intelligence.
Although the specific mechanism is unknown, the most important feature that distinguishes human intelligence from current artificial intelligence is that it is accumulated and developed through learning. Humans develop their intelligence through a long process of evolution, constantly accumulating what they see, hear, and feel as they are born and grow up. Even the process of picking up objects is easy because our bodies remember the repeated failures we experienced as infants, without the need for complex calculations. No matter how intelligent a robot may appear to be, if it is not capable of learning and developing on its own, but merely acts according to algorithms programmed by humans, it cannot be said to possess intelligence. The computer Deep Blue, which defeated the world chess champion, was not able to beat humans because it had a better understanding of chess than humans. Deep Blue was able to do so because it was able to list the countless possibilities of chess moves millions of times faster than humans, based on algorithms designed by humans based on the rules of chess. Just as calculators are far better at calculations than humans but cannot be considered intelligent, Deep Blue can beat humans at chess but is merely a machine that produces output based on given input and cannot be considered intelligent in terms of chess. To be considered intelligent, one must have the ability to understand and learn on one’s own.
According to evolutionary theory, humans evolved from single-celled organisms without intelligence about 3 to 4 billion years ago. We have reached our current state through repeated evolution from a state of obvious lack of intelligence. Considering that lizards appeared about 100 million years ago, the development of artificial intelligence is progressing at a rate thousands of times faster than that of living organisms. However, if the focus of AI technology is not shifted from “How can we process complex calculations at a faster speed?” to “How does the human brain learn, remember, and utilize information?”, the development of AI technology will gradually slow down and reach its limits. Of course, the human brain is not the only way to achieve human-level intelligence. However, the answer lies within us, so there is no need to search for other paths that are not yet clear.
