Just how smart is artificial intelligence?
16 July 2013
One of the best available artificial intelligence systems has been 'IQ-tested' to see how intelligent it really is.
Artificial and natural knowledge researchers at the University of Illinois at Chicago (UIC) have discovered that artificial intelligence (AI) is about as smart as the average four year-old, and reported their findings to delegates at the recent US Artificial Intelligence Conference in Bellevue, Washington.
The UIC team put ConceptNet 4, an artificial intelligence system developed at MIT, through the verbal portions of the Weschsler Preschool and Primary Scale of Intelligence Test, a standard IQ assessment for young children in the United States.
They found ConceptNet 4 has the average IQ of a young child. But unlike most children, the machine's scores were very uneven across different portions of the test, as report lead author, UIC's Professor Robert Sloan explains:
"If a child had scores that varied this much, it might be a symptom that something was wrong. But ConceptNet 4 did dramatically worse than average on comprehension the 'why' questions."
While ConceptNet 4 did very well on a test of vocabulary and on a test of its ability to recognise similarities, Professor Sloan said that one of the hardest problems in building an artificial intelligence is devising a computer program that can make sound and prudent judgement based on a simple perception of the situation or facts - in other words, common sense.
Common sense has eluded AI engineers because it requires both a very large collection of facts and what Professor Sloan calls 'implicit facts' - things so obvious that we don't know we know them. A computer may know the temperature at which water freezes, but we know that ice is cold, for example. Life is a rich learning environment; Professor Sloan again:
"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 common sense - that is, AI that can answer comprehension questions with the skill of a child of eight."
He and his colleagues now hope their study will help to focus attention on the so-called 'hard spots' in AI research.
Contact Details and Archive...