An AI system has beat humans on the verbal reasoning questions of an IQ test.
Originally shared by Wayne Radinsky
An AI system has beat humans on the verbal reasoning questions of an IQ test. These questions require understanding analogies, classifications, synonyms and antonyms. The system works by using a machine learning classification system to first classify the question as analogy, classification, synonym, or antonym, then use a "word embedding" system, which represents each word by a vector that represents how that word is related to other words, then, for each specific type of question, it will do something different and use a solver specific for that type of question. Multiple solvers were tested, using various algorithms including the Latent Dirichlet Allocation model, the Skip-Gram model, Multi-Sense model, and the Relation Knowledge Powered model.
http://www.technologyreview.com/view/538431/deep-learning-machine-beats-humans-in-iq-test
An AI system has beat humans on the verbal reasoning questions of an IQ test. These questions require understanding analogies, classifications, synonyms and antonyms. The system works by using a machine learning classification system to first classify the question as analogy, classification, synonym, or antonym, then use a "word embedding" system, which represents each word by a vector that represents how that word is related to other words, then, for each specific type of question, it will do something different and use a solver specific for that type of question. Multiple solvers were tested, using various algorithms including the Latent Dirichlet Allocation model, the Skip-Gram model, Multi-Sense model, and the Relation Knowledge Powered model.
http://www.technologyreview.com/view/538431/deep-learning-machine-beats-humans-in-iq-test
Comments
Post a Comment