When OpenAI revealed GPT-4 earlier this month, the company lauded its ability to “solve difficult problems with greater accuracy.” The language model passed the Uniform Bar Exam, LSAT, and USA Biology Olympiad with flying colors. As such, you probably won’t be surprised to learn that ChatGPT can also breeze through an IQ test without breaking a sweat.
ChatGPT scored a 155 on an IQ test
This week, Scientific American published an opinion piece by clinical psychologist Eka Roivainen in which he describes his experience administering the Wechsler Adult Intelligent Scale (WAIS) to ChatGPT. He opted for the third edition of the test, which consists of six verbal and five nonverbal subtests. Given that ChatGPT isn’t a physical being, it was only capable of completing the verbal subtests, which make up the Verbal IQ component of the WAIS.
Roivainen says five verbal subtests — Vocabulary, Similarities, Comprehension, Information, and Arithmetic — can all be presented in written form. The sixth verbal subtest, Digit Span, measures short-term memory, so he had to leave that one out.
He started with the Vocabulary subtest, which he had little doubt ChatGPT would crush.
“This subtest measures word knowledge and verbal concept formation, and a typical instruction might read: “Tell me what ‘gadget’ means,”” Roivainen explained. “ChatGPT aced it, giving answers that were often highly detailed and comprehensive in scope and which exceeded the criteria for correct answers indicated in the test manual. In scoring, one point would be given for a thing like my phone in defining a gadget and two points for the more detailed: a small device or tool for a specific task. ChatGPT’s answers received the full two points.”
This continued throughout the IQ test, with ChatGPT also receiving top marks in Similarities and Information. Its comprehension skills were admirable as well, and it was unsurprisingly capable of solving virtually any arithmetic problem thrown its way.
I kinda spoiled it above, but ChatGPT’s Verbal IQ was 155. That’s higher than 99.9% of the 2,450 test takers that make up the American WAIS III standardization sample.
ChatGPT has its limits, even as the language models powering it continue to evolve, but this is quite an impressive showing from a chatbot that will only get smarter.
GIPHY App Key not set. Please check settings