←back to thread

471 points doener | 1 comments | | HN request time: 0s | source
Show context
loloquwowndueo ◴[] No.45945033[source]
> The silhouette of Brown is neatly demarcated against the bubbling surface of the Sun. His downward trajectory is perfectly framed between sunspots, active regions on the surface of the Sun that are slightly cooler than their surrounding areas. This is not just a pretty picture; it is truly a masterpiece

Excuse me while I go wash off the stench of AI-generated descriptions. The picture is very nice, though.

replies(3): >>45957523 #>>45958597 #>>45960646 #
causal ◴[] No.45957523[source]
> This is not just a ...

Dead giveaway

replies(3): >>45958081 #>>45958508 #>>45959128 #
JumpCrisscross ◴[] No.45958081[source]
Kagi just launched an AI flagging feature. This is something HN needs.

I don’t mind AI content. But I’m not going to read it carefully before commenting, and will double check it with real sources before changing my mind about anything.

replies(3): >>45958894 #>>45958981 #>>45962826 #
squigz ◴[] No.45958894[source]
Has AI content detection become at all competent?
replies(3): >>45958947 #>>45959167 #>>45959365 #
1. codingdave ◴[] No.45959365[source]
From what I have seen, AI can detect AI better than humans. Humans are bringing baggage into it and vilifying legit writing techniques just because AI uses them. The whole concept that AI was trained on human writing, therefore uses well-known and proven writing techniques, but that those techniques are not exclusive to AI is utterly lost on some people.

The "dead giveaways" are not writing patterns, it is depth. AI will stay at a surface level when using argumentative writing patterns, whereas humans will add supporting information and connect the dots across sentences and paragraphs. It is the lack of connective language between thoughts and phrases that flag an AI.