You've most certainly already seen this image shared by Andriy Burkov and many others as being a great example of AI biases. To me this looks more like a good example of human biases. This algorithm is just showing what our society googles the most 🤷♀️ Notice I didn't use the plural here, that's because this is a statistical decision where the model only showed the top 1 or "best" result based on the data it was trained on (which is us, humans, googling). If it showed the "top 3" results, there would most certainly be a close fight shown between the probabilities of choosing "he", "she" or "it" for all of them. But using this top-1 result, it seems extremely biased but not because the data isn't representative of what we say. Rather because this is the problem, the data is representative, and the problem is most certainly us haha. It would be very hard to fix that, the only way would be to not represent the real world in my opinion. Or just change the real world, huh! 🤫 [this is just my opinion regarding this recent "trending post"] posted on Instagram - https://instagr.am/p/CM17Y5Gjvna/
