content/drafts/reducing-bias-and-improving-safety.md

   1 Title: "Reducing" Bias and Improving "Safety"
   2 Date: 2022-07-19
   3 Category: commentary
   4 Tags: artificial intelligence
   5 Status: draft
   6
   7 https://openai.com/blog/reducing-bias-and-improving-safety-in-dall-e-2/
   8
   9 > Today, we are implementing a new technique so that DALL·E generates images of people that more accurately reflect the diversity of the world's population. This technique is applied at the system level when DALL·E is given a prompt describing a person that does not specify race or gender, like "firefighter."
  10
  11 https://www.reddit.com/r/dalle2/comments/w1mflp/gender_bias_gone/igmevpj/
  12 https://twitter.com/waxpancake/status/1549076996935675904
  13 https://twitter.com/rzhang88/status/1549472829304741888
  14 https://twitter.com/cis_female/status/1549526889420947456
  15
  16 > it doesn't seem to want to give me a male nurse
  17 https://twitter.com/robertskmiles/status/1549559401622691840
  18
  19 https://www.lesswrong.com/posts/pRkFkzwKZ2zfa3R6H/without-specific-countermeasures-the-easiest-path-to
  20
  21 ------
  22