AI has an IQ across a 4-year-old

The first reviews of artificial intelligence (AI) in the form of human tests have been carried out. The results show that some AI systems have IQ equivalent to a 4-year-old.

AI system has intelligence by 4-year-old children

The team comes from the university. Illinois (USA) has tested the system A I with intellectual assessments of children. The results were reported by the MIT Review, the AI ​​system was completely defeated when the difficulty level was upgraded to 7 years old. At the same time, according to the affirmation, AI is currently only intellectually similar to a 4-year-old.

The AI ​​system involved in the research of the researchers is ConceptNet4 , a machine that scientists at university. Massachusetts (MIT), the US began to develop from the 90s of the 20th century, has so far developed along with many new technologies.

IQ tests are different in scope and form.

IQ tests are different in scope and form. They also depend a lot on the age of the respondents.

The IQ test is chosen by scientists named WPPSI (Wechsler Preschool and Primary Scale of Intelligence) . This test is often used at US schools.

This test is based on 5 classification criteria including:

1. Information - questions such as "Where can you find penguins?".
2. Vocabulary - questions such as "What is a home".
3. Reasoning ability - children will be provided with 3 suggestions such as "You can see through it", "It is square" and "You can open it".
4. The ability to understand problems - for example, "Why do we shake hands?".
5. Identify similarities - the example question is " Are pencils and pens all .?"

When asked if the questions of increasing complexity needed high thinking ability, ConceptNet4 could hardly answer the question.

The final assessment showed that the computer scored a high score on vocabulary ability and identification of similarities, which is quite good in processing information but inferior in reasoning and understanding ability.

The IQ test is chosen by scientists named WPPSI.

In some cases, researchers also have to suggest questions in different ways to test how ConceptNet4 connects data and responds to information. However, some questions are complicated and difficult to understand.

For example, "This animal has a mane will be a male" or "this is an animal that lives in Africa" and "this is a big yellow-brown cat". The AI ​​system can only return results listed by a list such as dogs, cats, farms, animals and families.

Or with the question "Why do we shake hands? ", The AI ​​answered with a list that includes a series of keywords like flirting, feeling or meeting friends.

Sometimes the answers are really absurd. The AI ​​responded to the results with words like "piano" and " band" for the question "Where can you find your teacher?".

Researchers cannot explain the unusual answers of AI. But they suggest that the results can be improved better if the question is entered through virtual assistant tools like Siri or Cortana.

The teaching methods for AI have started to grow amazingly.

The history of AI research began in the 1950s. In the early days, machines were provided with a knowledge base and operation based on logical thinking and reasoning.

However, the teaching methods for AI have started to grow surprisingly in recent decades. As science and technology develop more and more powerful, AI is absorbed more and more essence of human civilization. In particular, they can learn by themselves without the help of people.

Researchers believe that both approaches will be useful in the future. But the AI ​​can learn new knowledge by itself also poses potential dangers for human destiny itself. Can AI become smarter than humans and hold power.