A new preprint study reveals that OpenAI's GPT-4.5 model outperformed the Turing test, with participants mistaking it for a human 73% of the time. This study, which also evaluated other models like Meta’s LLama 3.1, reflects advancements in AI capabilities and suggests a substantial leap in AI’s conversational mimicry. Participants had difficulty distinguishing AI from humans, highlighting the evolving benchmarks for machine intelligence originally defined by Alan Turing in 1950. Lead researcher, Cameron Jones, pointed out that the AI often fooled participants even better than actual humans did.
In this study, researchers found that OpenAI's GPT-4.5 was identified as human 73% of the time in a Turing test format, far above chance.
Lead author Cameron Jones emphasized the striking ability of GPT-4.5 to masquerade as human, as it was often judged to be human over actual people.
Collection
[
|
...
]