Hacker News Reader

Comment by SwellJoe | Hacker News Reader

You're misunderstanding what these models do. It is a limitation of LLMs. They don't have memory, they do not learn, they cannot learn. The sooner you let go of your desire to have them learn or remember anything, the sooner you will achieve enlightenment (or, just a peaceful life where there is no possibility of getting into an argument with a machine).

If you want it to synthesize information that is not in its training data (from a few months ago), you can ask it to research the topic. But, arguing with an LLM is like putting lipstick on a pig. Only the machine is incapable of becoming annoyed. It has infinite patience to continue being wrong forever.

Your mental model of what Claude is and does is the problem here. Short of a revolutionary breakthrough in AI techniques, the LLMs will continue to do matrix math across a huge bunch of weights that cannot change based on anything you say.

View on HN

That's wrestling with a pig. "You both get dirty, and the pig likes it."

I guess putting lipstick on a pig might entail some wrestling, but it's a different idiom.

I was actually thinking of "Never attempt to teach a pig to sing; it wastes your time and annoys the pig."

So, yes, mixed up idioms. But, the machine doesn't like arguing either. It is incapable of liking or disliking things.

This is also a change in specifically Opus 4.8 / perhaps Fable 5 (I didn't really get enough of a baseline to see it there as much), where it's much more skeptical. For my purposes, this is fabulous - one of my pat addendums to most prompts is "challenge my assumptions and check the evidence empirically", and boy does it.

> fabulous

I think you mean fableuous ;)

They did not misunderstand anything. All of the behaviour is not inherent in raw base model and has been planted by the agressive, secretive reinforcement learning they do for benchmaxxing, "safety" and all other things. Claude begins any other sentence with "honestly". That is not how LLMs work, that is how they work after being RLed to the brink.

>Your mental model of what Claude is and does is the problem here. Short of a revolutionary breakthrough in AI techniques, the LLMs will continue to do matrix math across a huge bunch of weights that cannot change based on anything you say.

Sorry, but your mental model is wrong.

LLMs do matrix math across "a huge bunch of weights that cannot change based on anything you say", but the matrix math and results are informed (key concept here) by what you said, including the memory of what you said earlier in the discussion (and in some setups, even across discussions).

That's what a bloody prompt does.

It's entirely logic for the parent to want the LLM's matrix math + model + internal prompt, to accepts its prompt about LegalEagle and work with that, instead of arguing and giving him shit about it.

Especially since the earlier version of the model consistently worked like he wanted, and the new one consistently doesn't. He's not asking for some new unforeseen capability unknown to LLMs.

Exactly that.

I provided a question, and when given an incomplete answer, I provided with more info.

It refused to accept the additional info due to limited access to Youtube.

There was nothing more than that. There were no expectations.

The hostility and the amount of assumptions here are very strange.

...almost as strange as having a website accuse me of hallucinating a video and trying to gaslight it :D

You need to think this thought through all the way to the end. What it has said also influences what it will say. If it has consistently made combative responses, then the most likely thing to do is to continue to be combative.

I don't think there is any way back after the conversation takes a turn like that so there is no point in arguing with it. The only thing you can do is to fork the conversation before it made the first mistake and give it more context or tell it to look things up.

I wasn't arguing with it.

> "The only thing you can do is to fork the conversation before it made the first mistake and give it more context or tell it to look things up."

This is a key detail that many folks don't seem to understand about LLMs in general. The generation of a response happens based on the model weights and the context window (the system prompt + everything it's fed about the conversation thus far + any additional data included as part of the overall prompt). Each response technically stands alone and is generated entirely from only that context given to it and the model's existing "token space" weights. The illusion of an ongoing conversation is maintained "behind the scenes" by keeping that "context window" updated with the current state of the conversation as context for the next prompt, but the next response is technically an entirely new generation of text.

What it all means in a TL;DR sense is that the fix for a refusal is not to continue the "argument", but simply to remove that entire interaction from the conversation entirely as if it never happened and try a different tack with new / updated / more complete context to get the response you're expecting / seeking.

But unless you're using the API, it's not just a model.

I asked Gemini Flash 3.5 through the Gemini app something that followed a similar pattern. I asked about something, it replied with outdated info, I said that's outdated, it did a web search and apologized for being wrong, then proceeded to give me good info.

That wasn't just a bare model, that was a model wrapped in a harness, driving the model and allowing for web searches for example.

GPT in Codex is even more aggressive, I often see it proactively do web searches to ensure it's not feeding me wrong info.

You seem to be making a lot of assumptions about how I interacted in the messages to Claude.

You also seem to be making a lot of assumptions about my understanding of the models, especially considering I just told a story :)

I never said anywhere I want it to learn or remember, or that I argued with it.

I just provided additional information to it (in the form of a dozen or so words, tops, per message) and it accused me of hallucinating and trying to gaslight it.

My messages never went beyond a dozen words or so.

Show some examples, otherwise we're talking about interpretations.

I've already given enough.

I'm not gonna argue if you doubt it, I've been training argument dodging :)

Haha, would be a trip if this commentor is actually a Claude sockpuppet illustrating the point.

Yep haha. This happens quite frequently in HN, the famous [citation needed], so it might have been trained with data from here :/

No, I mean the actual prompt and its output. "I said this and it did that" is just a recall of your own memory, not an example. I don't want to argue with you, I'm interested in real stuff.

I swear I'm real :)

On the other hand, that's what a machine would say!

The machine is real, too!

Checkmate!!!

Having said that, since we are both real, I was seriously hoping to see some transcripts of one of such discussions.

I don't have it. I did it at work during lunch a few days ago so it's in incognito mode to not pollute the chats.

I thought nothing of it until I saw this discussion, so I saw no reason to save/screenshot.

It's ok if you don't believe in me.

Ah, it's not about believing, or not believing. I'm interested in Anthropic fumbles.

Claude?

Haha! I never considered the above message was parody, but it indeed mirrored that interaction perfectly!

Are you introducing yourself?

yeah yeah and human brains are just cells firing ions with small electric charge

very witty and very cynical, thank you

Reply on news.ycombinator.com