One of the most advanced artificial intelligence systems is about as smart as preschooler, new research suggests. But your preschooler may have better common sense.
An artificial intelligence (AI) system known as ConceptNet4 has an IQ equivalent to that of a 4-year-old child, according to the study. But the machine's abilities don’t evenly match those of a child: the computer system aced its vocabulary test, but lacked the prudence and sound judgment of a preschooler.
“We’re still very far from programs with common sense — AI that can answer comprehension questions with the skill of a child of 8,” study co-author Robert Sloan, a computer scientist at the University of Illinois at Chicago, said in a statement.
Artificial intelligence has been getting exponentially smarter for decades, and some technologists even believe the singularity — the point when machine intelligence will overtake humans — is near. Along the way, we’ve seen the trivia-playing computer Watson trounced trivia maven Ken Jennings at Jeopardy, and chess program Deep Blue eventually beat chess master Gary Kasparov in 1997.
But despite those high profile wins, AI's track record is decidedly spotty. Machines can't yet be programed to form intuitions about the physical world without doing extensive calculations, and they seem to fail at answering open-ended questions.
In the new study, researchers decided to test out just how close to human intelligence AI has come. They administered the verbal portions of a standard IQ test called the Wechsler Preschool and Primary Scale of Intelligence Test to ConceptNet4.
The machine's score was similar to that of a 4-year-old child. But this was no ordinary child: its intelligence was all over the map. The program scored highly on the vocabulary and similarity portions of the test, but floundered on comprehension sections, which are heavy on the "why" questions.
"If a child had scores that varied this much, it might be a symptom that something was wrong," Sloan said in a statement.
Not surprisingly, the computer also seemed to lack common sense. Through rich life experience, most people gather subconscious knowledge of the world that they rely on to make quick, wise judgments.
AI's "childhood" is positively deprived in comparison. For instance, AI may know the boiling point of water, but people know better the importance of steering clear of a hot stove.
"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. Life is a rich learning environment," Sloan said.
An artificial intelligence (AI) system known as ConceptNet4 has an IQ equivalent to that of a 4-year-old child, according to the study. But the machine's abilities don’t evenly match those of a child: the computer system aced its vocabulary test, but lacked the prudence and sound judgment of a preschooler.
“We’re still very far from programs with common sense — AI that can answer comprehension questions with the skill of a child of 8,” study co-author Robert Sloan, a computer scientist at the University of Illinois at Chicago, said in a statement.
Artificial intelligence has been getting exponentially smarter for decades, and some technologists even believe the singularity — the point when machine intelligence will overtake humans — is near. Along the way, we’ve seen the trivia-playing computer Watson trounced trivia maven Ken Jennings at Jeopardy, and chess program Deep Blue eventually beat chess master Gary Kasparov in 1997.
But despite those high profile wins, AI's track record is decidedly spotty. Machines can't yet be programed to form intuitions about the physical world without doing extensive calculations, and they seem to fail at answering open-ended questions.
In the new study, researchers decided to test out just how close to human intelligence AI has come. They administered the verbal portions of a standard IQ test called the Wechsler Preschool and Primary Scale of Intelligence Test to ConceptNet4.
The machine's score was similar to that of a 4-year-old child. But this was no ordinary child: its intelligence was all over the map. The program scored highly on the vocabulary and similarity portions of the test, but floundered on comprehension sections, which are heavy on the "why" questions.
"If a child had scores that varied this much, it might be a symptom that something was wrong," Sloan said in a statement.
Not surprisingly, the computer also seemed to lack common sense. Through rich life experience, most people gather subconscious knowledge of the world that they rely on to make quick, wise judgments.
AI's "childhood" is positively deprived in comparison. For instance, AI may know the boiling point of water, but people know better the importance of steering clear of a hot stove.
"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. Life is a rich learning environment," Sloan said.
0 comments:
Post a Comment
Grace A Comment!