[Window of science] artificial intelligence is also 'intelligence together'
- 자연과학대학
- Hit3235
- 2017-05-11
[Window of science] artificial intelligence is also 'intelligence together'
Kim Beomjun
Professor, Dept. of Physics, Sungkyunkwan University
There is much more to it than what you know, through what process a person's thinking takes place. However, it can be said that the neurons are gathered in the brain safely wrapped with a rigid skull and the information is processed through electrical signals by connecting them together. One nerve cell normally keeps the voltage inside it lower than outside. Then, when the sum of the currents flowing from other neurons is larger than a certain value, the voltage of the neuron is rapidly increased and then reduced to a negative value, as if the arrow fires rapidly. It is the 'fire' phenomenon of neurons. As stronger as they are aiming at the target by pulling tightly before the shot, the nerve cells use a large amount of energy to maintain a low voltage before ignition. In other words, our brain is an inefficient body organ that uses a lot of energy even when it is not processing information.
Ignited neurons transmit electrical signals to other neurons through long branches. There is a structure called 'synapse' between two neurons, which is known to store information in this area. The bigger way the nerve cell works is actually simple. The information sent by the other neurons is added to reflect the weights given to each synapse, and the total amount is then calculated to determine whether or not to ignite. An electrical signal that is ignited is output to other neurons and then input to the next neuron. In this way, it is not much different from simple electric device operation. It is a reason why we cannot comprehend the wonderful intelligence of a person to a single nerve cell. Although each operation is simple, the 'connection' of many neurons makes intelligence. A person's intelligence is also 'intelligence together’.
The contributions of physicists have been great in the development of artificial intelligence. In the artificial neural network of Hopfield, we simply describe the state in which the neuron is in a state of being ignited by +1 and the state when it is in a resting state quietly. It is similar to the spin of an electron with only two values: up and down. I have also made an artificial intelligence program that recognizes Hangeul characters. It was not new from the beginning. In physics research, we have already modified some of the programs we have already created and changed them into character recognition programs. It is interesting that physics’ finding of the energy floor of the spin system and recognizing the Korean alphabet using an artificial neural network are the same problem.
Deep-learning artificial neural networks nowadays are the same. One nerve cell is very simple to implement like the spin of physics, but like a real person's brain, it is a complex artificial neural network that collects a lot of these things. It builds neural networks into layers and learns the network as a person learns, realizing astonishing artificial intelligence like Alpha Go. Just as information learned by humans is ultimately imprinted on synapses that connect neurons in the brain, the learning of Go followed by a method of adjusting the weights of interconnects connecting artificial neurons. I am more wondering of the accomplishments and accomplishments of the accumulated scientists who have learned from the working principle of the human brain than by the skill of the Go of the Alpha Go, and have made this remarkable artificial intelligence. Artificial intelligence is also the result of man's learning from a man. However, scientists do not even know how the latest artificial intelligence created by humans works. As for the intellect of man, there are only a few fragmented clues at present, and no one still knows what process is recorded in information and rises to the surface of consciousness again. We learn something new every day, but we have not yet learned what it is to learn how to learn.
Source: http://news.chosun.com/site/data/html_dir/2017/04/04/2017040403568.html