@jwcph
Ironically enough, medical diagnosis was one of the first big applications of big AI. Anyone remember Watson AI and the computer that played on Jeopardy? Pattern marching on carefully trained data produced answers that were sound and could be curated by trained doctors. Note that someone that could understand and judge the result in my statement? It didn't make the bank IBM hoped for trust and privacy issues. Maybe doctors weren't happy with the idea of a tool that could replace them?
Generating and creating content from an indiscriminately trained AI isn't the same thing as a system like Watson. Because these systems mimics human creativity, people get suckered into thinking the results are the same as a human. They aren't. It's pattern matching and the AI judging the probability that the next word is the right word/meaning based on trained data. It's not understanding. Not even close. Alone, it can generate probable answers, but it takes a trained human to determine what's crap. Especially true if the answer is obscure—is it something good nobody thought of, or simply a technobabble hallucination?
@sfwrtr@jwcph Medical doctor here. It would be great if the computer could guess better than medical doctors what treatment could advert death and improve quality of life. Alas, the computer only could guess what selected doctors would have done, which is not exactly a gold standard
@sfwrtr@jwcph there's research about interpreting scans, biopsies etc for many decades now. Perhaps something is really useful?? Anyway, it's not commercially available when I practice, so I can't recommend patients to get their x-rays or whatever reported by an AI to save some bucks or be more accurate
Spot on. It even addresses why I love Node.js and hate NPM modules that programmers think are safe to use instead of coding their own. I haven't read to the end of the article, but the assessment in the beginning and why, especially about being a bubble is good.
Add comment