Assessing Political Bias in Language Models – Stanford HAI

Assessing Political Bias in Language Models – Stanford HAI

There are 250 billion such words behind GPT-3.5, the model fueling ChatGPT, for instance, and GPT-4 is now here.