[wookmaster]

If you keep hearing anecdotes at what point is it statistically important ? IBM 15 years ago was selling a story about a search engine they created specifically for the medical field(they had it on jeopardy) where doctors spent 10 years before they figured this poor patients issue. They plugged the original doctors notes into it and the 4th result was the issue they took a decade to figure out. Memorizing dozens of medical books and being able to recall and correlate all that information in a human brain is a rare skill to be good at. The medical system works hard to ensure everyone going through can memorize but clearly search engines/llms can be a massive help here.

[utopiah]

> If you keep hearing anecdotes at what point is it statistically important ?

Fair question but one has to keep in mind about ALL the other situations we do NOT hear about, namely all the failed attempts that did take time from professionals. It doesn't the successful attempts are not justified, solely that a LOT of positive anecdotes might give the wrong impressions that they are not radically most negative ones that are simply not shared. It's hard to draw conclusions either way without both.

[542354234235]

I hear about people winning the lottery all the time. There were two $100m+ winners just this week. The anecdotes just keep piling up! That doesn't mean the lottery is a valid investment tool. People just do not understand how statistically insignificant anecdotes are in a sufficiently large dataset. Just for the US population, a 1 in a million chance of something happening to a person should happen enough to be reported on a new person every weekday of the year.

[wookmaster]

The opposite is true also, if no company is paid or eager to dig into it, it could be significant and we don't know.

[ryandrake]

You guys are getting downvoted but you're 100% right. You never hear the stories about someone typing symptoms into ChatGPT and getting back wrong, bullshit answers--or the exact answer their doctor would have told them. Because those stories are boring. You only hear about the miraculous cases where ChatGPT accurately diagnosed an unusual condition. What's the ratio of miracle:bullshit? 1:100? 1:10,000?

[utopiah]

> You guys are getting downvoted but you're 100% right.

Classic HN. /s