“Our cheap text generator doesn’t have good content filters, so it’s transphobic” isn’t a “malfunction”, it’s operating as designed.
This concept that ML should ingest all the crap in the world, and then somehow fix it afterwards, is neither safe nor possible.
Follow
🇳🇱
I think everyone is looking at this the wrong way.. Its better to think of ML as a human being in a sense.. It will behave like whomever it learns from. If it learns from how other people in general behave, it will mimic that. If people are biased and racist the AI will be bias and racist... AI should not be seen as something that can be better than a human, it can just do what humans do, with all the flaws of a human.
