What is "machine learning"?

Machine learning works by feeding a system large amounts of example data and letting it gradually adjust internal parameters to recognize patterns, rather than a programmer writing out every rule by hand. Over many rounds of exposure and feedback, the system improves at tasks like recognizing images or generating text, effectively learning from experience the way humans get better through practice.

Robots assembling other machines describes physical manufacturing automation, unrelated to how software learns patterns from data, and memorizing the internet word for word mischaracterizes how these systems work, since they learn generalizable patterns rather than storing exact text.

Because machine learning systems learn from whatever data they're given, they can absorb and amplify biases present in that data, which is why training data quality has become as important a concern as algorithm cleverness.