Researchers claim that GPT-4.5 is the first large language module (LLM) to successfully pass the challenging three-way Turing test, convincing people that it is human 73% of the time. (Image credit: VLADGRIN via Getty Images)
Researchers report that large language models (LLMs) are becoming increasingly capable of imitating human skills, with GPT-4.5 now confidently passing the Turing test.
In a new study published March 31 on the preprint site arXiv, which has not yet been peer-reviewed, researchers found that GPT-4.5 was able to fool people into believing it was another person 73% of the time in a three-way Turing test. The study compared different artificial intelligence (AI) models.
While another group of researchers previously reported that GPT-4 passed a two-sided Turing test, this is the first time that LLM has successfully handled a more complex and original configuration of the “imitation game” proposed by computer scientist Alan Turing.
You may like
Sourse: www.livescience.com
