alt Hacker News> but I quickly concluded the writing suffered from the same uncanny valley effect as many AI-generated images: It all looks fine enough at first glance, but pay attention just a little longer, and something feels off.
My thoughts exactly. In all my interactions with gen AI it was always the same: on the surface it looks pretty convincing, but once you look more deeply it's obviously non-sense. AI is great at superficial imitation of human-created work. It fails miserably at doing anything deeper.
I think the biggest problem with AI is that most people just don't take the time or effort anymore to really look at an image, really read a text, or really listen to a piece of music, or a podcast. We've become so habituated to mindlessly consuming content that we can't even tell anymore if it's just a bunch of stochastic nonsense.
Replies
This is comparing LLMs to the best humans, and concludes that LLM output is "nonsense". Well, LLM output is better than the average human's output, and there are a many of humans at and below the average.
For four billion people, using an LLM to create things is a marked improvement. I'm not sure how you'd explain the phenomenally widespread use of LLMs otherwise.
By the way: Can you tell whether my comment (this one) was written by an LLM or not?
This is comparing LLMs to the best humans, and concludes that LLM output is "nonsense". Well, LLM output is better than the average human's output, and there are a many of humans at and below the average.
For four billion people, using an LLM to create things is a marked improvement.
Which if you think about it makes a lot of sense.
We’ve trained it so far on the outputs of our weird thinking process only.
https://www.astralcodexten.com/p/ai-art-turing-test
You can try to do a turning test. I've met several people claiming they can always find AI art, all of them can't do it (and AI art became even better now!)