#but decent enough to write and (more importantly) understand computer algorithms

[whyamiherewhosummonedme]

To elaborate some on how current machine learning algorithms inherently amplify bias, they are programmed to find patterns and repeat them. What they do not have, however, is reasoning for it. As it works now, AI will see that a disproportionate amount of the poor population in southeastern states is African-American and extrapolate from that fact. There is no understanding of the long history of oppression that caused this dynamic, only a recognition of its existence. This becomes a serious issue when you try to apply AI predictions to the real world, as you end up with the worst possible racial profiling. This is already a problem that has ruined hundreds of lives, with Palantir's machine learning software used by ICE to mark potential illegal immigrants - all of which were poor, working-class Latino families, and who were almost entirely released and aquitted because there was little to no actual proof, only the computer's profiling based on trends without taking into account the cultural nuances that lead to those trends.

AI can be useful. But you have to remember that it isn't thinking for itself. In its current form, it's pretty much a glorified search engine that, rather than giving you preexisting results, mashes them together to make one itself. It will always amplify biases, whether they be based on race, age, sex, disability, or more, and that is not a bug but the core of its very programming. Maybe in the future, we can create an AI that can indeed take external factors into account to fight this kind of inherent bias, but we don't have it now and everyone who uses or comes into contact with AI need to be aware of how exactly it works.

This is why it is so important to be critical and double check everything you generate using image generators and text-based AI.