IQ tests show that the AI Google one of them is the smartest, but it is still low by human standards.
The researchers who developed the IQ test for artificial intelligence, reported that Google is much superior to Siri and Bing, but has an IQ of less than six years of age.
According to researchers from AI Google has an IQ 47,28 on the basis of tests conducted during 2016.
The Google rating is a little below the IQ of six-year-old man of 55.5, but more than double the IQ of Siri to 23.9. According to them, Siri is not as smart as Microsoft Bing or Baidu who have the appropriate IQ and 31,98 32,92.
IQ score multiple search engines in 2014 showed that Google had a IQ of 26.5 compared with 23.5 Baidu.
"So far, the results show that the artificial intelligence system created by Google, Baidu and other had improved significantly over the past two years, but still have some gaps compared with the six-year-old child," the researchers write in an article published on ArXiv.
The researchers also propose a model to estimate the intelligence of AI systems and evaluating technologies in the context of the threat it poses to humans. Their model combines AI and human features around four areas of knowledge, including "input, output, skill and creation".
Model intelligence AI varies from system is first class which cannot exchange information with people under omnipotent godlike system of the sixth class, which is afraid of Elon Musk.
In many ways AI systems possess the same capacity for communication with people, like stones, while the system of the second level include smart televisions and washing machines, which can monitor information about the program, but do not have the ability to learn automatically.
The third class includes computers and mobile phones that can be programmed and upgraded, and the system of the fourth class Brain include Google, Baidu Brain and robots RoboEarth, thanks to the ability to adapt on the basis of information exchanged in the cloud.
People, in their opinion, can be seen to have "special" natural artificial intelligence that corresponds to the intelligence of the fifth class because of their ability to be creative, what is missing in all of the systems in the lower classes.
In this scheme, the researchers argue that AlphaGo from Google DeepMind will be system of the third class, partly because she is still at the stage of research and development. In addition, despite the victory over man in the game, it cannot be considered creative, because it still relies on people to teach.
"If people hadn't supported the program, and AlphaGo could get chess figures on the game, on its own initiative, to samoregulirovaniya and simulate the game to get experience to change your model of teaching to win in real competitions, we could say that AlphaGo can innovate," write the researchers.
"Because AlphaGo does not seem capable of such process, with a comprehensive point of view, its ranking is third class which is two class lower than that of men." |