Artificial intelligence is biased. Human beings are biased. In fact, everyone and everything that makes choices is biased, insofar as we lend greater weight to certain factors over others when choosing.
Developers design neural networks that can learn from data, but once they’ve released their creations into ‘the wild’, such neural nets have operated without programmers being able to see what exactly makes them tick. Hence, companies don’t find out their AI is biased until it’s too late.
Still, as much as AI has (deservedly) gained a reputation for being prejudiced against certain demographics (e.g. women and people of colour), companies involved in artificial intelligence are increasingly getting better at combating algorithmic bias.
Source: How Explainable AI Is Helping Algorithms Avoid Bias