This is one area where language models give us an interesting look at the question "What would English speaker right now assume about someone's gender from their name?" RoBERTa assigns an 88% probability to a Stevie being male, which tracks with the most famous Stevie (Wonder) and the near-exact 50/50 split between male and female Stevies in history. 98% of Stevies in 2021 were female, so that's definitely going to change, but for now Stevie is for many a boys name.
We can see similar patterns with other names: the AI lags behind the current data and changes less quickly for names with a long history (like Charlie as a nickname) or famous examples (Stevie). If you're naming a baby right now, the Social Security data is a better marker for how your child will fit in amongst their peers. For naming a fictional character or product, I think the language model is more helpful than demographic data, because it properly incorporates the things that influence how we unconsciously and consciously associate names with particular qualities.
This is one area where language models give us an interesting look at the question "What would English speaker right now assume about someone's gender from their name?" RoBERTa assigns an 88% probability to a Stevie being male, which tracks with the most famous Stevie (Wonder) and the near-exact 50/50 split between male and female Stevies in history. 98% of Stevies in 2021 were female, so that's definitely going to change, but for now Stevie is for many a boys name.
We can see similar patterns with other names: the AI lags behind the current data and changes less quickly for names with a long history (like Charlie as a nickname) or famous examples (Stevie). If you're naming a baby right now, the Social Security data is a better marker for how your child will fit in amongst their peers. For naming a fictional character or product, I think the language model is more helpful than demographic data, because it properly incorporates the things that influence how we unconsciously and consciously associate names with particular qualities.