A new artificial intelligence program has managed to pass a verbal IQ test aimed at young children.
To develop the software, Catherine Havasi and her team at MIT used a crowdsourced database filled with millions of statements that describe the relationships between everyday objects. The system ( known as ConceptNet ) drew upon this information to answer verbal questions such as "what is a house ?" that are common to IQ tests aimed at young children.
Impressively, the AI scored well in the areas of information, vocabulary and word reasoning, earning itself an IQ score in line with that expected of a 4-year-old. While the software was unable to tackle the spatial and symbolic reasoning parts of the test, future iterations may be capable of a far greater capacity for learning, verbal recognition and logic.
"Computers aren't really known for their way with words, but that could be about to change."
View: Full article | Source: New Scientist
Discuss: View comments (11)