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

Claude finds contradictions in my thinking

angadh.com

Post reply on HN
Print view View original

11–20 of 69 posts

Re: Claude finds contradictions in my thinking

#11

Post by Aurornis »

pendergast wrote: post #4

Fascinating. I feel like LLMs are great for the shy sections of society. They might hold beliefs, some strong, others weak. But they probably never speak these aloud for the fear of being judged. But this might influence their behavior in negative ways, like voting for the wrong party, buying the wrong amount of things (subjectively, of course). LLMs can act as a good foil here. Given enough context, they could iron…

From what I’ve observed, people are very good at getting LLMs to tell them what they want to hear.

Someone I know didn’t believe their doctor, so they spent hours with ChatGPT every day until they came up with an alternate explanation and treatment with an excessive number of supplements. The combination of numerous supplements ultimately damaged their body and it became a very dire situation. Yet they could always return to ChatGPT and prompt it enough different ways to get the answer they wanted to see.

I think LLMs are best used as typing accelerators by people who know what the correct output looks like.

When people start deferring to LLMs as sources of truth the results are not good.

Re: Claude finds contradictions in my thinking

#12

Post by lvl155 »

pendergast wrote: post #4

Fascinating. I feel like LLMs are great for the shy sections of society. They might hold beliefs, some strong, others weak. But they probably never speak these aloud for the fear of being judged. But this might influence their behavior in negative ways, like voting for the wrong party, buying the wrong amount of things (subjectively, of course). LLMs can act as a good foil here. Given enough context, they could iron…

I actually think it’s most useful for the extroverts of the society. The same people who speak before thinking. They need this to filter their thoughts and minds.

Re: Claude finds contradictions in my thinking

#13

Post by sandspar »

chrisweekly wrote: post #8

After reading this article ^1 about another writer's extreme disillusionment with using AI for feedback, I don't know if I'll ever trust it for this kind of thing. [1] https://amandaguinzburg.substack.com/p/diabolus-ex-machina

She's weirded out by creepy hallucinations, which is understandable! But ChatGPT is well known to hallucinate. In other words she doesn't know which of its behaviors are normal so she doesn't know how to react. Additionally, her particular issues are quite solvable with better prompting.

Re: Claude finds contradictions in my thinking

#14

Post by wongarsu »

bwestergard wrote: post #5
Lienetic wrote: post #2

Love the idea! It's hard to get an "unbiased" outside perspective, especially on more personal, inner thoughts. Will definitely try this out, thanks for sharing.

The difficulty is that these models will reflect the aggregate worldview of people on the web before 2022 or so.

If only that was true. ChatGPT has gotten a bit more subtle since the early days when it was allowed to criticize certain politicians but not others, but so-called "safety training" still seems to impart plenty of additional bias. Some others like Grok appear to be less biased, until they suddenly turn into mecha-hitler after a slight prompt tweak

Re: Claude finds contradictions in my thinking

#15

Post by ceejayoz »

achierius wrote: post #10
mtalantikite wrote: post #6

> You seem to be trying to integrate Eastern philosophy’s emphasis on acceptance with Western ideals of active engagement and personal agency. Where is the contradiction here? Westerners too often think of acceptance and compassion as this soft deference in the face of adversity. It's not! Sometimes compassion is wrathful, as we're talking about waking someone up to the true nature of what a situation is. As long as…

[flagged]

That particular deity appears to embody "nature isn't super concerned with compassion/fairness". https://en.wikipedia.org/wiki/Mahakala; "because they are Kala or Time in the personified form, and Time is not bound by anything, and Time does not show mercy, nor does it wait for anything or anyone"

That said… https://en.wikipedia.org/wiki/Skin_grafting

Re: Claude finds contradictions in my thinking

#16

Post by bgwalter »

I have not had any LLM correct me even once. In any three-prompt LLM session I find at least one logical mistake, one invalid correction on the part of the LLM and one piece of outright misinformation. All of which is delivered in an authoritative manner.

The only useful feature is that LLMs like Grok can find obscure tweets, but I just want the link like in a search engine, not a summary.

Re: Claude finds contradictions in my thinking

#17

Post by ceejayoz »

sandspar wrote: post #13
chrisweekly wrote: post #8

After reading this article ^1 about another writer's extreme disillusionment with using AI for feedback, I don't know if I'll ever trust it for this kind of thing. [1] https://amandaguinzburg.substack.com/p/diabolus-ex-machina

She's weirded out by creepy hallucinations, which is understandable! But ChatGPT is well known to hallucinate. In other words she doesn't know which of its behaviors are normal so she doesn't know how to react. Additionally, her particular issues are quite solvable with better prompting.

> If I do poor work with an electric drill then it's not the drill's fault.

> ChatGPT's sycophancy crisis was late April.

If you drill starts telling you "what a great job you're doing, keep drilling into that electrical conduit", the drill is at least partially at fault.

A tool that randomly and unpredictably fails is a bad tool. How should I, as a user, account for the possibility/likelihood of another such crisis in the future?

Re: Claude finds contradictions in my thinking

#18

Post by bwfan123 »

There is no right or wrong on the output of AI. You lapped it up wholesale because you were looking for meaning, and found it in the AI.

I have to invoke the BS generator here:

https://sebpearce.com/bullshit/

I can bet lots of people find profound meaning in the output of the above machine because they want to.

Post reply on HN