The more I think about this tweet, the angrier I get. This does not help your case @TwitterComms. This says not only did your training set and the team creating this model lack sufficient diversity, but apparently so did the team testing it. What makes me angry however is 1/
https://twitter.com/TwitterComms/status/1307739940424359936
that this is not the first, tenth or even hundredth time we've seen models fail like this. It's talked about widely in the tech space. Yet somehow @Twitter ignored all the discussions about how to avoid these biases and instead went with pre-delivery testing as the solution 2/
to eliminating such biases? And @dantley this isn't about dunking on you or @Twitter, this is about frustration at the pattern of behavior by tech giants who utilize ML and AI without being able to sufficiently address the inherent biases they introduce. It's a disturbing 3/
trend that continues to play out. The only silver lining here is you've given us yet another high-profile example of why #diversity isn't just a social issue, but also imperative to business success. I truly hope you can make meaningful changes to #DoBetterBeBetter. /FIN
• • •
Missing some Tweet in this thread? You can try to
force a refresh
