AI Machine Has Same IQ As Four-Year-Old Child

Researchers have once again tested the limits of AI by putting…

Increasingly machines are getting better at specific tasks such as playing chess, recognising pictures and making complex commutations.

Thanks to films such as The Terminator, many of us are a little spooked by computer “artificial intelligence”.

It explains that while it has taken 60 years to produce a machine with reasoning capabilities similar to a human child, “exponential improvements” over the next six years could see “dramatic improvement”.

But general intelligence is still proving elusive for a lot of them.

Stellan Ohlsson at the University of Illinois and his colleagues gave the standard IQ test to ConceptNet 4, a powerful AI machine built at the Massachusetts Institute of Technology (MIT).

An AI system has been shown to score higher for verbal IQ than a four-year-old.

The team administered the Verbal IQ (VIQ) part of the Wechsler Preschool and Primary Scale of Intelligence (WPPSI-III) to the ConceptNet 4 AI system. What the test tries to do is assess the ability of someone to rationally understand the world around them – it’s in this particular area of self-awareness where software is still a few way behind.

The AI machine did well on vocabulary and similarities, examples of questions in these categories were “What is a cat?” and “Rain and snow are both made of…?” The test questions (for example, “Where can you find a penguin?” and “Why do we shake hands?”) were translated into ConceptNet 4 inputs using a combination of the simple natural language processing tools that come with ConceptNet together with short Python programs. Commenting on the results he said: “ConceptNet does well on Vocabulary and Similarities, middling on Information, and poorly on Word Reasoning and Comprehension”. And its answer was “epileptic fit”. However, researchers have now quantified just how intelligent machines could be.

The history of AI research stretches back to the 1950s. That method of teaching computers has been surpassed in recent decades by a new era of machine learning, where AIs are given large quantities of data from which they can learn. A very important advance in AI since then is the switch from knowledge gathering to being learning-driven. Perhaps knowledge bases that are a hybrid of the two paradigms will play a role in the next round of AI progress.