I consider AI like human thinking but more prone to errors

[PoetPhilosopher]

AI doesn't remove prejudices but rather exposes the prejudices of the programmers, and often you don't get the same peer review within a computer program as when a human shares their opinion. Then add to that the additional layer of unintended machine learning errors and programmer errors. And then add to that that sometimes the programmers programming may not always have the best environment, or at least one that promotes good nutritional health and a good work environment and not hard pressed for deadlines on software bent on trying to change lives.

Sometimes I get tired when people, not on here, just in the world, try to godify AI.

 

[Debater Slayer]

AI learns from the datasets and patterns that are input into it by applying specific algorithms to them. It doesn't express the prejudice of the programmers so much as the prejudice of whatever data it learns from.

Let's say that, for instance, you have a dataset of rejected CVs and a dataset of accepted ones for a given job. You now want to train your AI by feeding it these datasets so that it learns how to pick out job applicants.

If the datasets include results of prejudiced hiring practices, such as exclusion of specific ethnic or religious groups, the AI will pick up this pattern and replicate it. It has no consciousness or moral compass, after all: it's a mere tool that learns from the inputs we give it.

Treating AI as a perfect solution to everything rather than a highly useful tool for many applications is mistaken. No tool is perfect, especially not one that is quite open to continual improvement (through machine learning) and optimization. AI has been significantly useful for various industries and industrial processes (such as automated assembly lines), but make no mistake: without proper human oversight and tuning, it can get derailed and become useless.

 

[Heyo]

AI doesn't remove prejudices but rather exposes the prejudices of the programmers, and often you don't get the same peer review within a computer program as when a human shares their opinion. Then add to that the additional layer of unintended machine learning errors and programmer errors. And then add to that that sometimes the programmers programming may not always have the best environment, or at least one that promotes good nutritional health and a good work environment and not hard pressed for deadlines on software bent on trying to change lives.

Sometimes I get tired when people, not on here, just in the world, try to godify AI.

Computers are much less prone to errors. Their hardware is much more rigid and not influenced by random quantum evens or equally random chemical imbalances. Thus, they will, within their ability, do exactly what you want them to do.

Be careful what you wish for - it might come true.