When people say
#AI and technology are racist, THIS is what we're talking about. Here is proof that the data used to train
@midjourney has led to a racist tool because of inherent bias that exists on the internet (this is just a symptom of a much bigger, very well documented problem within the training data they - and many
#LLMs - use).
I submitted a photo of a *very dark black* coworker sitting on a couch to see how it would handle the prompt of creating a headshot, and to my *utter* amazement, this is what it spit back out.
There are obviously far more critical situations where embedded bias in technology can drastically affect people's lives, such as in
#healthcare and law enforcement. This is just a very blatant, easily understandable example of how early we (as a society) are in solving for these data bias issues that have downstream affects on the quality of the AI output.
#ResponsibleAI development is critical if we're actually going to see the benefits of these tools. Otherwise we're just exacerbating these biases, some of which we may not even be aware of yet.