15 comments

[ 2.9 ms ] story [ 47.0 ms ] thread
In before 50 sexist comments bc HN doesn't do articles talking tech from specifically female perspectives.
I must be completely unimaginative because I've yet to find any AI model remotely interesting, or willing/able to talk about interesting things.

At the moment it feels like everyone in the world has gone insane except for me.

Have you tried GPT-4? What kind of interests do you have?
Honestly some of us cannot find a conversation with a bot interesting on premise. And not even because of its conversations capabilities or the weird moralistic limits they all have. I just cannot relate to a bot, even if it sounds incredibly human.

There is nothing interesting a bot has to say. It has no personality, but also has every personality at once. There’s no curiosity as to what it’s thinking because it’s not thinking. It has no preferences or dislikes or ability to relate to anything fundamentally human. I don’t get it, and it doesn’t matter how good bots get. At best it’ll be a tool but it’s never going to be my friend.

Very strange to base your views on the hidden internals about a thing rather than the experience of interacting with it.
Very human, why should I trust some matrix-multiplication and gradient descent on deeply personal questions. For sure that guy in the mirror looks amicable, but you don't get much out of it when trying to flirt with a mirror.

I have to try out GPT-4, but I don't think it will be able to help me with my hobby. Like getting the feel of a painting right, whereas my friends have an easy time pointing out flaws or give recommendations beyond some technical irrelevancy

I don’t know, why do you assume human minds are anything more than that?

The more I interact with an LLM the more it reminds me of my manager.

Correct me if I'm misunderstanding, you're putting forward the notion that the human mind is reducible to matrix-multiplications and gradient descent.

This seems like such a strong claim, that I shouldn't be the one arguing for it. Rather you should bring strong evidence.

Tbh,I agree with on the manager anecdote (maybe more on the cynical side)

I have interacted with it, and I find it incredibly outputting. Very strange you assume everyone would enjoy the experience.
No, it sucks chunks. But the reasoning here is still bad.
I have, and while it's impressive technologically it's just boring.

>What kind of interests do you have?

I've tried talking to it about all sorts of things. Music, electrical engineering, religion, culture, etc. It's just too agreeable to be interesting, and every time I think I'm getting close to getting it to disagree about something, it shuts down.

Also the tone is unmistakable, and something I haven't been able to shake. But it's like I could predict its responses before it writes them (again, definition of boring).

LLMS are basically very advanced parrots; I'm not surprised you find them boring.
Wow AI has really gotten realistic hasn't it
I tried one or two (free) ones when trying to figure out what AI can do. I wasn’t trying to get company but to see the tech’s limits. I made all sorts of weird unhinged characters and made them do all sorts of unhinged things.

Basically, they could be goaded into accepting anything. They were absurdly easy to gaslight and had no opinion or initiative of their own. Some AIs did a bit better, but none did well.

It left me worrying about the sorts of behaviours one would develop from interacting too much with AI and not enough with real humans. The things speak and act like humans, but feel lobotomised and submissive. They interact differently, even though they speak the same.