Overheard two dads at soccer practice arguing about whether AI will replace human proofreaders

One guy said he ran 500 pages through ChatGPT and found 12 errors it missed, which makes me wonder if we're putting too much trust in AI for stuff like that or if the other dad was right that it's just a tool that needs supervision, has anyone else tested it on a big project and found similar issues?

2 comments

2 Comments

wyatt_ross2725d ago

My buddy works in publishing and they tested an AI proofreader on a 1950s novel reprint. It flagged the phrase "Indian giver" as a potential error because it didn't understand the historical context was the point. That's the real problem, it can't tell when a "mistake" is actually important. It just looks for patterns, not meaning. So yeah, it'll miss stuff that needs a human brain to figure out.

the_mason25d ago

Wait, it tried to "correct" a phrase that was written on purpose? That's wild. So it would basically strip out anything that makes old writing feel real. How's it supposed to learn context if it just deletes anything it doesn't get? Feels like a fancy spellchecker that makes bigger mistakes.