GPT-3 has remarkable abilities in some areas, but is still not quite on the same level as a human in others. In his experiment "Giving GPT-3 a Turing Test", Kevin Lacker investigated GPT-3's abilities to determine how close the model is to passing a Turing test. He found that GPT-3 could answer simple factual questions, such as “Who was president of the United States in 1955?”, without any difficulty. However, it was also happy to answer nonsensical questions without recognizing that they made no sense.
This experiment assessed GPT-3's abilities in the following areas:
To learn more, read the article: https://lacker.io/ai/2020/07/06/giving-gpt-3-a-turing-test.html