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

Ask HN: How do you deal with people who trust LLMs?

news.ycombinator.com

Post reply on HN
Print view View original

201–210 of 236 posts

Re: Ask HN: How do you deal with people who trust LLMs?

#201

Post by basilikum »

lovelearning wrote: post #182
Jensson wrote: post #176

Earlier quoted context omitted.

> OP implies human sources are the "good sources" or "reputable sources." No OP did not do that, nowhere did OP mention human sources. When I search the internet I don't just find human sources, I also find automatically generated data graphs and maps and such, those are also good sources of data. If I had to choose between a map from Google maps and a map from an LLM I'd trust the map from google maps any day.

> automatically generated data graphs Are you saying there are data graphs that don't have humans in the chain? If so, what came up with the data and the tools to generate those graphs? And how do you decide which data and graph to trust? What exactly makes them "good sources"? > If I had to choose between a map from Google maps I would too. But Google Maps relies on local 3d party survey companies that use people, m…

> I feel my point that all these "reputable sources" or "good sources" have biases (and mistakes) still stands.

That does not contend with the premise. The premise is not that reputable sources are perfect or form objective truth.

The premise is that when you do not have the time to investigate something closer reputable sources are much more likely to be close to the truth than LLMs that are trained on these reputable sources as well as unreputable sources and just mimick these training source material thereby obscuring the source of information and introducing hallucinations.

When you have the time to investigate information, primary sources and reputable secondary sources allow you to more easily trace information and judge its validity. LLMs by their nonpredictable nature hinder this.

Thesis: "A is better than B"

Argument: "A is flawed"

The argument does not engage with or weakens the thesis.

Re: Ask HN: How do you deal with people who trust LLMs?

#202

Post by wolvoleo »

throwaway444422 wrote: post #141
wolvoleo wrote: post #85

At work I had this kind of discussion on a conference call, someone looked something up about a internal company policy and it came back with a hallucinated wrong result. So I said, don't ever trust the output of an LLM without verification. However this caused me some hassle with the AI adoption manager. We have minimum-use AI KPI's for employees and he asked me to stop saying these things or people will use it less…

If the tool provides a link to the raw source, you can just click that. The LLM is basically just a better search engine here.

Well yes, that's what I was trying to point out.

But that's not what anyone does. They just grab the blurb that comes out and paste it as an answer. People are lazy.

The same with the meeting minutes. Every time people just grab them from teams and mail them on without even reading it. Of course being full of errors (especially because teams is very very bad at determining who is speaking when speakers are on the same account (e.g. if multiple people are in a physical room with a table microphone).. And assigning tasks to the wrong person as a result.

I keep pointing these mistakes out but people just shrug and point to the little disclaimer: "AI generated content, may contain errors". Well yeah, that was a warning for YOU to doublecheck it before you pump that shit out.

They often respond also like: "Nobody reads the minutes anyway". Well yes but if they don't then why do we have them in the first place? Not doing it at all is even more efficient than using AI for it.

But anyway I've given up on fighting against this, I just ignore these people unless a task is incorrectly assigned to me.

Re: Ask HN: How do you deal with people who trust LLMs?

#203

Post by wolvoleo »

kombookcha wrote: post #128
wolvoleo wrote: post #85

At work I had this kind of discussion on a conference call, someone looked something up about a internal company policy and it came back with a hallucinated wrong result. So I said, don't ever trust the output of an LLM without verification. However this caused me some hassle with the AI adoption manager. We have minimum-use AI KPI's for employees and he asked me to stop saying these things or people will use it less…

That sounds insufferable and I'm sorry your coworker and your bosses have gotten this hypnotized. There are actual use cases, but this type of productivity cargo cult drives me up the wall. At least if they were into filling the office with magic crystals, it would be decorative and easily ignored. This is just forcing people to spend a bunch of tokens in a dull ritual to make the line go up.

Yes :( It is awful. Thank you.

Re: Ask HN: How do you deal with people who trust LLMs?

#204

Post by basilikum »

heliumtera wrote: post #133
basilikum wrote: post #76

Earlier quoted context omitted.

Most NPC like take. Look I get your sentiment. Sometimes it feels like you're the only thinking, conscious being. Surrounded by beings who fundamentally cannot understand that A –> B does not imply B –> A. Beings that say things that are so obviously non-sequiturs or contradictory. But calling people NPCs is the most NPC thing you can do. There is more to people than logical reasoning and these things often impede or…

I liked this response and overall I don't disagree. Just to be clear, I don't have any problems with people considered to be of lower intelligence. This is besides the point. Unfortunately, this isn't something that can be elaborated much further. Since you took the time to respond, and seems to be a little perturbed, I will just say again I was never addressing lack of intelligence or mental capabilities. Intelligen…

Intelligent ability is not the same as logical abstract reasoning in behavior. I think we agree on that. But intelligence is a prerequisite for it.

> They are the most likely to believe in absurd nonsense.

I disagree. Intelligent people who believe in absurd nonsense are better at rationalizing that nonsense and at making it convincing. Some people write very lengthy books arguing for a clearly flawed premise precisely to distract from the fact that the premise is flawed.

But overall intelligent people are less likely to believe in nonsense in the first place and more likely to critically and, importantly, logical soundly examine a thesis that is presented as given.

> Intelligent people convinced themselves, day one at school, there is a better way described on the books. They got really good when they understood this. They got pretty rewards.

What I think you are describing is academic (and professional, maybe somewhat social) performance and conformity, not intelligence.

Your answer reads increasingly convoluted to me towards the end. But I will try to interpret a sentiment from it.

People who do not arrive at the same conclusions as you do not necessarily do so for a lack of reasoning. Two equally intelligent individuals when presented with the same messy, incomplete fragments of reality can through reasoning arrive at widly different conclusions. And in reality we are never even presented with the same fragments of reality in the first place.

Man, if you think you have figured it all out, you really haven't. Regardless of what it is that you believe. There really is nothing more foolish or more NPC like than to think you've understood it all.

Re: Ask HN: How do you deal with people who trust LLMs?

#205

Post by dalke »

lovelearning wrote: post #193
dalke wrote: post #188

Earlier quoted context omitted.

Well, in my case going to a Peanuts comic and looking at hands was pretty easy, and didn't involve any questions about negative environment or labor consequences, the massive hammering of web sites to gather data, centralization of power, and the like. Like, "!w Peanuts" in my search bar, look at the image, and count fingers. "a rather outlier example" You wrote that you use AI to find "obscure connections" - aren't…

Even as a human, I find whatever sources Google shows to be inconsistent. I can't give any confident answer about the number of fingers. I think the answer is actually "4 sometimes and 5 other times." So I'm not sure how much LLMs can handle this kind of inconsistency between "reputable" visual sources and text sources, nor how representative this example is. A "reputable source" like Variety says this... https://var…

Your first link was cited in the 2nd half of Google AI's answer, and one of the top Google answers, so I think you are using Google as your information source.

The large majority of the images you link to show kids with 5 fingers, as well as 5-fingered baseball gloves. The cases of four fingers are due to orientation.

Your "1." also shows Marcie with five fingers. You see Charlie Brown with 4 fingers because he's holding a baseball. In 2. he's also holding a baseball. You would not see 5 fingers on one side because doing so would look strange.

In your unlabeled "0." there are plenty of kids with 5 fingers. There are some with fewer, but they are holding things or drawn in way to suggest we are seeing the hand from the side.

I don't understand your hesitancy. Your own samples should be enough for you to decisively conclude that the Google AI's claim that Peanuts was "traditionally drawn with four fingers (or three fingers and a thumb) on each hand" is wrong. If not, it sure seems like you trust Google AI over your own eyes. Why are you so hesitant to agree?

My point is that you don't need to consult secondary sources when the primary sources are easily available.

When this came up a few days ago, I spot checked the complete works of Peanuts, from a collection on archive.org at https://archive.org/details/peanutscomics19502000/Volume%201... . The consistent pattern across the nearly 50 years of Peanuts is the kids have five fingers unless obscured by orientation or objects.

You can do that yourself, and triple-check that Google AI's answer is clearly wrong.

Thus, I think it's a good example of how fact checking with LLMs can lead people astray, and the large negative externalities I mentioned combined with its well-known tendencies to make incorrect statements make it a very poor starting point when the primary source, at least in this case, is so easy to access.

If most of the sources are wrong, and LLMs are being trained on those, isn't it logical that the latter will also likely output that same wrong information?

When do you know if most of the sources are wrong, unless you yourself know most of the sources are wrong?

Re: Ask HN: How do you deal with people who trust LLMs?

#206

Post by esalman »

lovelearning wrote: post #196
esalman wrote: post #192

Earlier quoted context omitted.

> But a redeeming quality is that we can ask the same LLM to fact check its own answer step by step in real time with little effort. They often identify their own hallucinations and reduce the probability of retaining that mistake in the rest of the conversation. It is not really a redeeming quality of LLM. It falls apart because if you want to properly fact check an LLM, you have to make yourself aquinted on human-g…

Gemini (currently) gave this reasonable and unbiased answer that challenges both pro-US/Israeli and pro-Iranian biases. It's also a good point for the OP - when selecting an LLM, ask the people they're dealing with to ask 2-3 LLMs. It's much faster (and cheaper!) than reading about a complex geopolitical issue (that may not matter to them enough to spend a lot of money on books and research papers) while at the same…

What I'm hearing is that if you are dealing with something that matters to you a lot, you should avoid relying on LLMs.

Re: Ask HN: How do you deal with people who trust LLMs?

#207

Post by lovelearning »

dalke wrote: post #188
lovelearning wrote: post #179

Earlier quoted context omitted.

> When do you stop the fact checking? Exactly the same calculus as fact checking anything else from any other source. What are the social/economic/ethical consequences to me if the answer is wrong or inaccurate or incomplete? How much time do I have to check? How thorough should I be? I imagine this calculus isn't really that different for most people. Or is it? As for your example, I believe it. But I also feel it's…

Well, in my case going to a Peanuts comic and looking at hands was pretty easy, and didn't involve any questions about negative environment or labor consequences, the massive hammering of web sites to gather data, centralization of power, and the like. Like, "!w Peanuts" in my search bar, look at the image, and count fingers. "a rather outlier example" You wrote that you use AI to find "obscure connections" - aren't…

HN isn't showing me a reply option for your latest comment, so I'll reply here instead.

Just to clarify, I used plain Google search not Google AI mode. And opened search results which seemed "reputable," without knowing anything much about Peanuts cartoon or cartooning.

I had no idea at all about archive.org having it and didn't see it listed in the first two pages of search results.

I still find it confusing, especially given what the Variety.com link says which doesn't mention orientation. If the acceptable explanation for 4 vs 5 is orientation, why is it wrong when the AI generated 4 fingers? Does it not match the rest of the orientation?

Anyway, I'm not sure where this leaves LLMs. I'll explore image capabilities when I get some opportunity and keep your comment in mind.

Re: Ask HN: How do you deal with people who trust LLMs?

#208

Post by quirkot »

phito wrote: post #126
mathgladiator wrote: post #91

Earlier quoted context omitted.

Yes, but I am full in on simulation hypothesis, and people are going to enter the matrix... willingly. https://nexivibe.com/intj.html

While some of the ideas in this do resonate with me (or at least they're entertaining), it's unfortunate that's it's so obviously LLM generated. And some parts of it, like the INTJ exceptionalism, reek of LLM sycophancy, which then turned into to some kind of god complex...

observation a: Document title is about a minority's rightful supremacy

observation b: document says "this is not political" then dives into persuasive speech

conclusion: this document was written by the bad guys

Re: Ask HN: How do you deal with people who trust LLMs?

#209

Post by ryandvm »

ddawson wrote: post #13

I'm going to hold them to the same standard no matter if they use crappy sources, plagiarize, or hallucinate on their own. If someone asked, when and if I am in a position where I have to tell them, I would remind them that LLMs prioritize their own confidence over correctness. LLMs aren't a special case to me. Glue doesn't belong on pizza and you shouldn't eat one rock a day but we've been giving and getting bad adv…

Yeah, it is not clear to me that the average person is going to do any better with googling then they are with asking an LLM. At least the LLM is mostly the average of all published knowledge (and misinformation).

I feel like people that ask questions like this must have much smarter friends and family members than I do.

I know people that still believe in Pizzagate or chemtrails or that vaccines cause autism. Clearly finding reputable information sources is not a strong suit for a lot (half?) of the population.

Re: Ask HN: How do you deal with people who trust LLMs?

#210

Post by eudamoniac »

beeflet wrote: post #113
eranation wrote: post #21

Ask them to tell the LLM it's wrong... then when it goes "You are absolutely right!" to challenge it and say that it was a test. Then when it replies, ask it if it's 100% sure. They'll lose faith pretty quick.

I tried to fool claude sonnet with confidence and it failed. https://claude.ai/share/47145af0-47d1-451b-813c-131ec48e7215 Maybe it is possible with a more complex or subjective question.

I could not fool chatgpt free version with this same line of questioning.
Post reply on HN