Many of us wonder how artificial intelligence came about. Artificial intelligence is an attempt to recreate human capabilities. At the same time, artificial intelligence languages do not build logical models of functioning. Rather, they look at the world in the context of other words.
Artificial intelligence is not the work of a genius who has suddenly been dazzled. We simply have more data and better equipment. This huge amount of data is a magical component of artificial intelligence. Data we didn't have access to before. Our civilization has created a technology that allows us to capture and store massive amounts of data.
There are billions of gyro, accelerometer, GPS, fingerprint, tweet and photo smartphones in the world.
Cameras installed in cars absorb information about the world around us. Buoys submerged in seas and oceans provide data for climate models. Toaster-sized satellites observe infrared life. The whole world is flooded with this type of data.
What were the beginnings of artificial intelligence? Not very media oriented. Most people don't realize that Google is a form of artificial intelligence. You enter a question into the search engine and receive an answer. Each search improves the algorithm. Sometimes the answers appear before we finish typing the question. And we already have. We don't even have to go to the target page. As you remember in the past it wasn't that easy.
In 2010 and 2011, the Watson machine created by IBM and named after its creator began to win the Jeopardy quiz show on a regular basis! Watson turned out to be faster than human players. Watson's machine read Wikipedia and 200 million pages in natural language. Nobody thought that the machine was fast enough and precise enough to win over a man. The machine learned by reading. Watson trained on text algorithms but did not understand it. He does not know that fire can burn something. He sees the world in a completely different way than we do. But it is Watson who performs business analyses today, for which analysts used to be paid 300 USD per hour. Watson's victory in the tournament has gone down in the history of artificial intelligence.
The last decade of rapid development of artificial intelligence was based on pattern recognition and machine learning. It is a process based on algorithms that learn from data, recognizing dependencies and regularities without relying on traditional man-made software. But this has changed in recent years. The so-called deep learning has appeared.
It's all about the machines learning by themselves. We call this "deep learning" and it means the same as a neural network. It is a new approach. A neural network is a simulation of brain function. We have long nerves in our brain that connect with thousands of others. They create paths in connection with the tasks to be performed. When a child manages to do something successfully, the path is strengthened. Otherwise, the path weakens. In this way, the child improves his or her interaction with the environment. The computer learns like a child. It gets a lot of data and draws its own conclusions. He is wise as he gains experience.
Within the Google project, a neural network test was carried out.
Imagine a situation in which 1000 computers suddenly wake up and know nothing. The programmers made the whole system watch youtube for a week, so he was learning about gangnam style movies and other everyday stories. Then an analysis was made. What turned out to be? Of course, a neuron responsible for detecting human faces was created. The second neuron was used to detect cats.
A few years ago, Deep Mind became a semi-dependent company of Google. The company's goal is to create a digital superintelligence more powerful than the intelligence of all the people together.
One of the systems supporting Deep Mind's teaching wakes up like a newborn baby. It only sees the game screen on the ATARI computer. He knows nothing about objects, movement or time. He sees only the screen and the result. Deep Mind goes through every game. Passes all ATARI games at an inhuman pace. It ends in less than a minute. If a child on the day of birth mastered 40 ATARI games on a superhuman level, we would consider them possessed.
Another challenge for Deep Mind was playing Go. It was thought that the computer would never defeat a man in this game. Deep Mind faced the European champion first. It was October 2015. He won. Then he faced the Korean champion in March 2016 and won. People started to comfort themselves that Go was just a board game.
But there is still poker. Poker is an art that requires people to read. We are dealing with bluffing. This is not science. In 2017 the best poker players were gathered in Pittsburgh. After a week of gameplay, the computer demolished them all. As you can see, artificial intelligence needs time to acquire a new skill, but when it does it is unstoppable.