World's Best Artificial Intelligence System is as Smart as a Four-Year-Old

A new study on the accuracy of one of the world's best artificial intelligence (AI) systems has shown that it is as smart as a four-year-old. However, the AI system lacks commonsense, researchers said.

The system chosen for the study was the ConceptNet 4, an artificial intelligence system developed at M.I.T. Researchers from University of Illinois at Chicago put the system through an IQ test called 'Weschsler Preschool and Primary Scale of Intelligence Test'. Although the system "thinks" like a four -year-old child, its answers are very uneven. Researchers say that if an average child answered the questions so unevenly then the child would be deemed to having a neurological disorder, meaning that the system has some developmental problems.

The system obtained good score in vocabulary and can find similarities. However, it could not answer the "why" questions, Robert Sloan, professor and head of computer science at UIC, and lead author on the study, said in a news release.

In 2011, a team of researchers at University of Götheberg reported that they had created a computer program that scored 150 points in an IQ test. But the catch there was that the IQ test was designed to suit the computers' ability. Many scientists have been trying to get computers to think like a human- to make simple logical connection based on large amounts of data.

"All of us know a huge number of things. As babies, we crawled around and yanked on things and learned that things fall. We yanked on other things and learned that dogs and cats don't appreciate having their tails pulled. We're still very far from programs with commonsense-AI that can answer comprehension questions with the skill of a child of 8," said Sloan.

The results of the current study will be presented, July 17 at the U.S. Artificial Intelligence Conference in Bellevue, Washington.

Researchers in the 50s and 60s had much faith in the concept of AI , but by the 70s and 80s this faith started to wane. Even now, researchers can't teach computers the basics of common sense. In an article posted in 2012, Yusuke Miyao, an associate professor at the National Institute of Informatics (NII), Japan, talks about how difficult it is to teach computers the implicit knowledge about the world that humans acquire naturally.