Hacker News Reader

Comment by threethirtytwo | Hacker News Reader

That twitter post was clearly written by AI along with the instructions for the AI to avoid "tells" and other tropes common to AI.

Absolutely nothing wrong with something written with AI. Just pointing it out.

View on HN

There is something wrong with humanity losing the willingness to think and type out a four-sentence paragraph, I would wager.

Generated comments are banned on HN, FWIW.

But was it written with an OpenAI AI?

Must've been.

We're nearing or at an inflection point where people like this are dependent on it.

if that's the case, ai failed to remove the negative parallel construction (my current top ai smell aside from slanted inverted commas). what signs are there of this being ai asked not to sound like ai?

Right, that's the sign. Ai often fails to do what it's told. So that's the sign of it asked to not sound like an AI. I told AI to do this for my current post as well.

ok. do you see any more concrete signs? to me it smells like openai output with newlines removed. but aside from smell (and the negative parallel construction), one could argue that this may be the output of a human who has been influenced by the prose of ai.

Perhaps there isn't really a functional difference.

as a black box there evidently is no functional difference, as output is perceptibly identical in both cases. as a white box its amusing to consider that an ai worker may themselves become verbally indistguishable from ai, although in this particular case its more likely that the ai worker is simply lazy and told ai to write a tweet for them.

and you say this based on?

The way everyone else can tell. My instincts. AI has a flavor.

Reply on news.ycombinator.com