Detecting Hate Speech with GPT-3

• DOI: 10.48550/arxiv.2103.12407
• Publication Date: 2021-01-01
• Authors: Ke-Li Chiu; Annie Collins; Rohan Alexander
• Keywords: Zero (linguistics)

Sophisticated language models such as OpenAI's GPT-3 can generate hateful text that targets marginalized groups. Given this capacity, we are interested in whether large be used to identify hate speech and classify sexist or racist. We use racist passages with zero-, one-, few-shot learning. find zero- one-shot learning, an average accuracy between 55 per cent 67 cent, depending on the category of type With model's high 85 cent. Large have a role play detection, further development they could eventually counter speech.