Excellent article, Yennie!

Expand full comment

This is a fantastic overview! I wish I could have read it before concluding my recent (independent) research in this area:


Very interesting to see your mention regarding the more pronounced gender bias toward generating male historical figures.

In the above-linked work, I show how time and gender are, in fact, correlated in LLMs (ranging from BERT to GPT-4 (preview) Turbo), with a proposed causal relationship described in the work. I look forward to reading your piece on “Where are all the women?” to see if you have similar findings.

Expand full comment

Sorry, realized the GPT-4 results were updated in arxiv only: https://arxiv.org/abs/2210.00131

Expand full comment