AI is an attempt to help computers appear to think. A simple example is matching parts of an object to the object itself. Humans are good at that, but computers are real bad at that.
For example, if I showed the average human 50 pictures of parts of a dog, such as a tail, a snout, an ear etc, even a child can identify it as a dog. A non-AI computer would not know what it is, as they need to match the pictures to a whole dog to ID it. If the computer just happened to have the exact picture of the dog part stored and labeled as a dog, it could ID it. That is the basis of AI.
What the data centers do is take small pieces of data, say a dog's ear and label it as a dog, and ask the computer to match it to the picture. They need to do this with each and every piece of data, hence the huge amount of processing and storage requirements.
nOf course, if they mislabel, or misidentify the data, mistakes are them made in identifying the item.