The departure of AI ethics researcher Timnit Gebru from Google has led to a lot of discussion and debate in the AI and AI ethics communities. If you are just tuning in, here is what happened: AI ethics researchers at Google (including Gebru) wrote a research paper on the dangers of large language AI models, in collaboration with academic researchers from the University of Washington. (GPT-3 from OpenAI and BERT from Google are examples of large language models.) According to this paper, these models have environmental costs-they need massive computing infrastructure and consume large amounts of electricity-and they can normalize racist and/or sexist language, because they are trained on a corpus of internet text that includes such language. In a nutshell, Google wanted the paper retracted because it contested the claims in it. Gebru did not comply, and this led to her exit from Google. (She was unceremoniously forced out, although Google maintains she was not explicitly fired.)
展开▼
