AI, right now, is the latest technical phrase being abused for marketing purposes. It's what happens when Google and Microsoft realize they've taken search engines about as far as they can go. They're labeling it with a new gimmick, stuffing it into everything that'll sit still, and hoping you'll notice it.
In reality, AI is a poorly understood label. In its most direct form, AI is composed of two computers (to reduce it to simple terms). The first computer gets fed a bunch of data. You tell that computer "this is what I'm looking for. When you see these patterns, I want you to do this thing. It's meaningful to me". Once you give that computer a sufficient amount of data to train on (you think it knows the pattern you're wanting and not others), you let that computer talk to the second one.
The second one is where the "real world" data is being fed. In the beginning, it's sorting through all that stuff looking for the patterns it was told about. When it thinks its found one, it gives you a result. You tell the second computer if it was right or not. It uses that information to update itself and the first computer. If done correctly, the two computers work together to get better and better at realizing what you want and rejecting what you don't want.
Of course, left to the public, a malicious operator will poison the results by telling the computer that results praising Hitler or sexually graphic descriptions of children are ideal. The public is stupid and immature.
Done properly, AI is a godsend for things like seeing deep patterns in cancer cell genomes, predicting protein folding outcomes, the natural progression of traffic patterns, and other complex iterative models.
You don't need it to find an Italian restaurant nearby.
