Dunn acknowledges that he could have found the same answer with the right Google search terms, but says that the point is that he didn’t have to: ChatGPT immediately returned what he was looking for even though he described it vaguely.
I remember when google used to return the right results even when the search was vague.
Its so stupid, google “bipartie matching algorithm” and the second result is a stack overflow where the second answer is the Hungarian algorithm…
So every programmer would have found that immediately using the traditional methodology…
Yup, AI is frequently a worse way to do something we used to do, but with a crappy pretend to be a person responding veneer.
That’s not what I got. In fact, stack overflow isn’t in any of my results. I got a lot of scholarly articles for various algorithms and none of them mention the Hungarian algorithm.
Weird, google might know I only click on stackoverflow links lol
https://stackoverflow.com/questions/23059735/fast-maximum-matching-algorithm-for-bipartite-graphs
Same lol. Its incredible how shit it’s been made.
Right? I was thinking that exactly. I remember being in awe by how well google could do that, and now I actually dread having to use it and sift through all the crap.
I guess the same will happen to “AI” once they try to monetize it with ads.
The question he asked to ChatGPT doesn’t seem particularly vague to me. He used various algorithm names and concepts such that all he really asked is “I want a fruit like an apple but not as round” and it responded with “pear”
Find a new algorithm? It’s not creative, it just gave up one it knew about.
Imagine an article for TF1:
A valve engineer used Google to find a new matchmaking algorithm for Team fortress and now it’s in the game
