learninglinguist: Link to the original tweet  I’m happy the whole thread is here, and tha

learninglinguist:

Link to the original tweet 

I’m happy the whole thread is here, and that it correctly notes that these aren’t explicitly programmed in by a single mustache-twirling product engineer but instead they reflect larger systemic biases as a result of how the algorithm works — likely an algorithm that seemed neutral while being planned out of context.

This reminds me of the various AI chatbots that became racist and sexist when they interacted with the public. AI algorithms optimize for what they see as most common and most successful in the environment they’re in. (I wish there were a follow-up experiment to see what personalities those chatbots would have ended up with if they were set loose on various gated or self-moderated communities instead of the internet at large.)

The work you do and the things you create must exist inside the world; you must take that into consideration during the act of creation. Every choice has meaning.