i'm not a big openAI fan and even I think the title is crummy. I think it should be "How we used chatGPT to make a 10k mistake." -- at least then it's honest a

i'm not a big openAI fan and even I think the title is crummy. I think it should be "How we used chatGPT to make a 10k mistake." -- at least then it's honest about the party at fault, that being the startup that didn't vet generative code. Relatedly i've been throwing AIs at the problem of ordering mods and dependencies for game engines; it's pretty astonishing the error rates you see involving medium-sized text lists. A good experiment: take a 100 line text file of whatever, ask an AI to sort the lines by some criteria and output a text file, you'll get files back with less than 100 lines routinely. These kind of things really limit my faith in those systems to work without a heavily leashed supervisor along side.