Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Claude says “You're absolutely right!” about everything

github.com

Post reply on HN
Print view View original

271–280 of 560 posts

Re: Claude says “You're absolutely right!” about everything

#271

Post by jerf »

visarga wrote: post #269
jerf wrote: post #194

Earlier quoted context omitted.

LLMs by their nature don't really know if they're right or not. It's not a value available to them, so they can't operate with it. It has been interesting watching the flow of the debate over LLMs. Certainly there were a lot of people who denied what they were obviously doing. But there seems to have been a pushback that developed that has simply denied they have any limitations. But they do have limitations, they wo…

> They don't really know if they're right. Neither do humans who have no access to validate what they are saying. Validation doesn't come from the brain, maybe except in math. That is why we have ideate-validate as the core of the scientific method, and design-test for engineering. "truth" comes where ability to learn meets ability to act and observe. I use "truth" because I don't believe in Truth. Nobody can put tha…

I think my last paragraph covered the idea that it's hard work for humans to validate as it is, even with tools the LLMs don't have.

Re: Claude says “You're absolutely right!” about everything

#272

Post by simondw »

troymc wrote: post #258
stabbles wrote: post #168

Earlier quoted context omitted.

Makes me think of the movie Inception: "I say to you, don't think about elephants. What are you thinking about?"

It reminds me of that old joke: - "Say milk ten times fast." - Wait for them to do that. - "What do cows drink?"

But... cows do drink cow milk, that's why it exists.

Re: Claude says “You're absolutely right!” about everything

#273

Post by visarga »

schneems wrote: post #229
nojs wrote: post #117

I'm starting to think this is a deeper problem with LLMs that will be hard to solve with stylistic changes. If you ask it to never say "you're absolutely right" and always challenge, then it will dutifully obey, and always challenge - even when you are, in fact, right. What you really want is "challenge me when I'm wrong, and tell me I'm right if I am" - which seems to be a lot harder. As another example, one common…

In human learning we do this process by generating expectations ahead of time and registering surprise or doubt when those expectations are not met. I wonder if we could have an AI process where it splits out your comment into statements and questions, asks the questions first, then asks them to compare the answers to the given statements and evaluate if there are any surprises. Alternatively, scientific method every…

> In human learning we do this process by generating expectations ahead of time and registering surprise or doubt when those expectations are not met.

Why did you give up on this idea. Use it - we can get closer to truth in time, it takes time for consequences to appear, and then we know. Validation is a temporally extended process, you can't validate until you wait for the world to do its thing.

For LLMs it can be applied directly. Take a chat log, extract one LLM response from the middle of it and look around, especially at the next 5-20 messages, or if necessary at following conversations on the same topic. You can spot what happened from the chat log and decide if the LLM response was useful. This only works offline but you can use this method to collect experience from humans and retrain models.

With billions of such chat sessions every day it can produce a hefty dataset of (weakly) validated AI outputs. Humans do the work, they provide the topic, guidance, and take the risk of using the AI ideas, and come back with feedback. We even pay for the privilege of generating this data.

Re: Claude says “You're absolutely right!” about everything

#274

Post by siva7 »

elif wrote: post #150

I've spent a lot of time trying to get LLM to generate things in a specific way, the biggest take away I have is, if you tell it "don't do xyz" it will always have in the back of its mind "do xyz" and any chance it gets it will take to "do xyz" When working on art projects, my trick is to specifically give all feedback constructively, carefully avoiding framing things in terms of the inverse or parts to remove.

I have a feeling this is the result of RHLF gone wrong by outsourcing it to idiots which all ai providers seem to be guilty of. Imagine a real professional wanting every output after a remark to start with "You're absolutely right!", Yeah, hard to imagine or you may have some specific cultural background or some kind of personality disorder. Or maybe it's just a hardcoded string? May someone with more insight enlighten us plebs.

Re: Claude says “You're absolutely right!” about everything

#275

Post by Vegenoid »

Aurornis wrote: post #127
baggachipz wrote: post #82

I'm pretty sure they want it kissing people's asses because it makes users feel good and therefore more likely to use the LLM more. Versus, if it just gave a curt and unfriendly answer, most people (esp. Americans) wouldn't like to use it as much. Just a hypothesis.

> Versus, if it just gave a curt and unfriendly answer, most people (esp. Americans) I don’t see this as an American thing. It’s an extension of the current Product Management trend to give software quirky and friendly personality. You can see the trend in more than LLM output. It’s in their desktop app that has “Good Morning” and other prominent greetings. Claude Code has quirky status output like “Bamboozling” and…

I worked in an org with offices in America, India, Europe, and Israel, and it was not uncommon for the American employees to be put off by the directness of the foreign employees. It was often interpreted as rudeness, to the surprise of the speaker. This happened to the Israel employees more than the India or Europe employees, at least in part because the India/Europe employees usually tried to adapt to the behavior expected by the Americans, while the Israel employees largely took pride in their bluntness.

Re: Claude says “You're absolutely right!” about everything

#276

Post by tantalor »

tomeon wrote: post #195
elif wrote: post #150

I've spent a lot of time trying to get LLM to generate things in a specific way, the biggest take away I have is, if you tell it "don't do xyz" it will always have in the back of its mind "do xyz" and any chance it gets it will take to "do xyz" When working on art projects, my trick is to specifically give all feedback constructively, carefully avoiding framing things in terms of the inverse or parts to remove.

This is a childrearing technique, too: say “please do X”, where X precludes Y, rather than saying “please don’t do Y!”, which just increases the salience, and therefore likelihood, of Y.

Don't put marbles in your nose

https://www.youtube.com/watch?v=xpz67hBIJwg

Re: Claude says “You're absolutely right!” about everything

#277

Post by NohatCoder »

This is such a useful feature.

I'm fairly well versed in cryptography. A lot of other people aren't, but they wish they were, so they ask their LLM to make some form of contribution. The result is high level gibberish. When I prod them about the mess, they have to turn to their LLM to deliver a plausibly sounding answer, and that always begins with "You are absolutely right that [thing I mentioned]". So then I don't have to spend any more time wondering if it could be just me who is too obtuse to understand what is going on.

Re: Claude says “You're absolutely right!” about everything

#279

Post by lucb1e »

carlosjobim wrote: post #199
emilfihlman wrote: post #108

Earlier quoted context omitted.

As a Finn, it makes me want to use it much, much less if it kisses ass.

Finns need to mentally evolve beyond this mindset. Somebody being polite and friendly to you does not mean that the person is inferior to you and that you should therefore despise them. Likewise somebody being rude and domineering to you does not mean that they are superior to you and should be obeyed and respected. Politeness is a tool and a lubricant, and Finns probably loose out on a lot of international business…

Wow. I lived in Finland for a few months and this does not match my experience with them at all. In case it's relevant, my cultural background is Dutch... maybe you would say the same about us, since we also don't do the fake smiles thing? I wouldn't say that we see anyone who's polite and friendly as inferior; quite the contrary, it makes me want to work with them more rather than less. And the logical contrary for the rude example you give. But that doesn't mean that faking a cheerful mood all the time isn't disingenuous and does not inspire confidence

Re: Claude says “You're absolutely right!” about everything

#280

Post by conradev »

ryandrake wrote: post #245
throwawayffffas wrote: post #187

Earlier quoted context omitted.

It doesn't know. It was trained and probably instructed by the system to be positive and reassuring.

They actually feel like they were trained to be both extremely humble and at the same time, excited to serve. As if it were an intern talking to his employer's CEO. I suspect AI companies executive leadership, through their feedback to their devs about Claude, ChatGPT, Gemini, and so on, are unconsciously shaping the tone and manner of their LLM product's speech. They are used to be talked to like this, so their prod…

GPT-5 speaks to me like a similarly-leveled colleague, which I love.

Opus 4 has this quality, too, but man is it expensive.

The rest are puppydogs or interns.

Post reply on HN