This is actually true.

Most notably to me, the ability to sift through and collate enormous amounts of data has led to surprising things like diagnosing diabetes through retinal scans.

But those sorts of things, beneficial and impressive though they might be, remain at the fringe of AI research for the simple reason that those sorts of uses are too niche to provide the revenue stream that all of the bubble-building corporate parasites demand. Their focus is on the AI-as-a-substitute-for-real-intelligence aspect (and increasingly “AI” as just a meaningless marketing buzzword), since that’s where the money is. And unfortunately but not coincidentally, that’s where most of the public attention is too.