Main menu




Can superintelligence be born?

[Science Questions Solved with Cartoons] The Future of Artificial Intelligence

Sci-fi movies sometimes show that in the future, artificial intelligence that far surpasses human intelligence will become self-conscious and destroy humanity. So, people have a vague fear of artificial intelligence. Could artificial intelligence really be born in the future that could destroy mankind?

Artificial intelligence network system and deep learning

Artificial intelligence is a computer program that realizes what human intelligence can do. When artificial intelligence was first developed, it was a program in which artificial intelligence stores intelligence-related rules and information in a computer, and the computer makes decisions by itself by comparing it with information received from outside. So, there were many programs that put a large amount of knowledge in a specific field and made the computer to make judgments or reasoning like an expert.

However, this artificial intelligence was far from real human intelligence because it could not do things that humans could do very easily. For example, no matter how much information it had about a cat, it couldn't tell the difference between a cat and a dog by looking at a picture, and couldn't do trivial things like distinguishing colors.

However, starting in the late 1980s, scientists started making artificial intelligence programs using artificial neural network models that mimic the learning process of the human brain. The most popular machine learning method using artificial neural networks these days is deep learning. In general machine learning, the computer learns in advance and classifies data based on it, but deep learning does not learn in advance, but discovers patterns in numerous data input to the computer, learns by itself, and classifies data just as humans classify objects.

Therefore, the larger the amount of data, the more effective deep learning is possible. For example, in the existing machine learning, supervised learning was required to teach a cat by inputting a large amount of pictures of cats in order for a computer to distinguish between cats. However, in deep learning, without such a supervised learning process, a computer can learn to identify cats by analyzing numerous cat pictures. The reason AI is capable of such deep learning is that computers, the Internet, and mobile devices have become necessities of life, and the enormous amount of text and image data poured through them can be used by computers.

Thanks to deep learning, the current artificial intelligence has developed to the level of finding the same person in another photo or video by analyzing the facial features of a person in a photo, and it has become able to understand conversations that are not grammatical by learning the conversations of many people.

Can strong artificial intelligence and superintelligence be born?

As artificial intelligence learns by itself through deep learning, people are very interested in how far artificial intelligence will develop. People think that if artificial intelligence is highly developed, it can develop to a level where not only human intellect and reason, but also emotions and self-consciousness knowing that it is an artificial intelligence can be developed. Such artificial intelligence is called strong artificial intelligence, and furthermore, artificial intelligence that far surpasses human intelligence is called 'superintelligence'. It is this strong artificial intelligence and superintelligence that is destroying mankind in science fiction movies. Could such a strong artificial intelligence really be born in the future?

Many experts argue that the birth of strong artificial intelligence is impossible in the near future. In order to create strong artificial intelligence, it is necessary to know the entire function of the human brain, but the research is still in its infancy. The human brain has 100 billion brain cells called neurons, and there are about 100 trillion synaptic connections between those neurons.

Since synaptic connections are so complex and the process of transmitting information involves more than 100 types of neurotransmitters as well as electrical signals, no one knows how long it will take to fully understand the functions of the human brain. Google's artificial intelligence required a supercomputer with 16,000 central processing units to distinguish a cat's face, but it may not be possible no matter how many central processing units are connected to imitate human intelligence. Moreover, experts predict that the performance of computers will soon reach its limits.

However, the birth of strong artificial intelligence is not at all impossible. Scientists are looking forward to a quantum computer that uses the characteristics of quantum mechanics in which 0 and 1 exist at the same time, unlike the existing digital computer that processes only 0 and 1 signals. This is because quantum computers can perform calculations millions of times faster than conventional supercomputers. There are still many hurdles to overcome to create a perfect quantum computer, but using a quantum computer may enable the birth of strong artificial intelligence.

On the other hand, if you watch the movie Terminator, a complex connected Internet network suddenly becomes self-conscious, and Skynet, a strong artificial intelligence, is born. Can artificial intelligence suddenly increase its intelligence like this? When many computers are connected to each other, the phenomenon in which a computer's intelligence suddenly increases without input of new information is called 'emergency of intelligence'. In fact, some argue that the emergence of intelligence is possible in computer science. However, most experts say that these claims are greatly exaggerated. It's like the odds of making a 747 airliner by accident while repairing a wagon.

If strong artificial intelligence is born, can artificial intelligence rule humans as in science fiction movies? Experts' opinions are divided. There are opinions that warn of the dangers of strong artificial intelligence, while others believe that humans can control sufficiently strong artificial intelligence. Also, there is an opinion that there is no reason to attack humans no matter how self-conscious the AI ​​is. It is argued that artificial intelligence cannot generate desire or greed because there is no body that constantly feels deficient. Rather, the human desire to exploit the power of artificial intelligence that has become more powerful for the worse could be a bigger problem.