Artificial intelligence is a long way to go? It is difficult to have 10 years old children's common sense resolving power.

According to foreign media reports, for humans, the concept of common sense is relatively easy to identify, although it is difficult to define accurately. For example: standing at the end of the crowd at the time of ticket purchase; people holding the ends of red-hot metal firearms, so as to avoid being Burns; these common senses seem simple, but they are not simple for artificial intelligence machines.

How do we teach these vague common sense of artificial intelligence robots? Many researchers tried to do this but eventually failed. But it is likely that this will change soon. At present, Microsoft co-founder Paul Allen Participate in this project study.

According to the New York Times, Allen is investing $125 million in his non-profit computer lab, the Allen Artificial Intelligence Institute (AI2), and plans to double the investment budget over the next three years. The funds will be used For existing projects, and the 'Alexander project', the latter is a new initiative focused on teaching robots 'common sense concepts'.

Allen pointed out at the press conference that when I created the AI2 Institute, I hope to expand the capabilities of artificial intelligence robots through high-impact research. In the early stages of artificial intelligence research, people have paid much attention to the concept of common sense, but this Work remains stagnant. Artificial intelligence robots still lack the common sense concept that most 10-year-olds have. We hope to initiate this research and achieve major breakthroughs in this area.

If the robots are very advanced, they can simulate human tasks, they can locate and identify objects, climb, sell houses, provide disaster assistance, etc.

However, even these advanced robots cannot handle simple problems and instructions. How do they respond to an unusual situation and how to use 'common sense' to correct proper behavior and response? Now they still cannot be realized.

Oren Etzioni, executive director of the AI2 Institute, said: “Despite the recent success of artificial intelligence, it is still difficult for artificial intelligence to distinguish between them, but this is very simple for humans. There is currently no artificial intelligence system that accurately answers a series of simple questions. For example: If I put socks in a drawer, will it be there tomorrow? Or how do you know if a bottle is full?'

Ezioni stressed that, for example: When AlphaGo artificial intelligence program defeated the world’s number one chess player in 2016, AlphaGo did not know that go is a board game.

One of the simple reasons why we don't teach the concept of AI common sense is: The concept of common sense is actually very difficult to understand. Gary Marcus, founder of Geometry Intelligence, is inspired by common sense and abstract thinking in child development, London. Imperial College researchers focus on artificial intelligence symbolic technology, which is a technique for human beings to mark anything with artificial intelligence.

So far, no research strategy has helped robots identify 'common sense concepts'. The Alexander Project will take a harder approach to solving this problem. It will integrate research into machine reasoning and computer vision and find a common sense of measurement. Methods.

Gary Marcus, the founder of Geometry Intelligence, said at the press conference that I was very interested in the Alexandrian project. It is time to solve this problem in a new way. But if artificial intelligence develops to the next At the stage, and combined with more aspects of human life, we must solve the common sense concepts of robots. The current Alexander project may be the best attempt.