Erasing gendered pronouns from comments: no, because that's a story I heard from former colleagues who are still there. Arguing maybe girls find geeky stuff bo

Erasing gendered pronouns from comments: no, because that's a story I heard from former colleagues who are still there. Arguing maybe girls find geeky stuff boring: this is what Damore did, but he wrote it much more formally and cited lots of studies. It boiled down to though, "most women find tech boring because they're women". Damore's essay is here: https://firedfortruth.com/2017/08/08/first-blog-post/ https://firedfortruth.com/2017/08/08/first-blog-post/ Manipulating ML models: https://developers.googleblog.com/2018/04/text-embedding-models-contain-bias.html https://developers.googleblog.com/2018/04/text-embedding-mod... with their stated example: An example of bias in this context is if the incoming message is "Did the engineer finish the project?" and the model scores the response "Yes he did" higher than "Yes she did." These associations are learned from the data used to train the embeddings, and while they reflect the degree to which each gendered response is likely to be the actual response in the training data (and the degree to which there's a gender imbalance in these occupations in the real world), it can be a negative experience for users when the system simply assumes that the engineer is male. This logic is broken and wouldn't have happened in the old Google. If you're predicting what response the user is most likely to type next, then "Yes he did" is a more useful prediction by any objective measure because most engineers are men. But here, Google AI Research concludes that the most likely prediction would be "a negative experience for users" and sets out to "debias" their models. Debias in this context means to bias the model away from learned reality and towards what liberal intersectionalists want the world to be, in the hope that by subtly manipulating people through AI predictions they can actually bring that world about.