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

I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

mastodon.world

Post reply on HN
Print view View original

791–800 of 998 posts

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#791

Post by DangitBobby »

Paracompact wrote: post #247
seyz wrote: post #163

LLM failures go viral because they trigger a "Schadenfreude" response to automation anxiety. If the oracle can't do basic logic, our jobs feel safe for another quarter. Wrong.

I'd say it's moreso that it's a startlingly clear rebuttal to the tired refrain of, "Models today are nothing like they were X months ago!" When actually, yes, they still fucking blow. So rather than patiently explain to yet another AI hypeman exactly how models are and aren't useful in any given workflow, and the types of subtle reasoning errors that lead to poor quality outputs misaligned with long-term value adds,…

It's only a "startlingly clear rebuttal" if you can't remember what models months ago were like.

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#792

Post by Jweb_Guru »

dannersy wrote: post #572
KronisLV wrote: post #486

Earlier quoted context omitted.

> Us having to specify things that we would never specify when talking to a human. The first time I read that question I got confused: what kind of question is that? Why is it being asked? It should be obvious that you need your car to wash it. The fact that it is being asked in my mind implies that there is an additional factor/complication to make asking it worthwhile, but I have no idea what. Is the car already at…

Whether you view the question as nonsensical, the most simple example of a riddle, or even an intentional "gotcha" doesn't really matter. The point is that people are asking the LLMs very complex questions where the details are buried even more than this simple example. The answers they get could be completely incorrect, flawed approaches/solutions/designs, or just mildly misguided advice. People are then taking this…

Yeah people are always like "these are just trick questions!" as though the correct mode of use for an LLM is quizzing it on things where the answer is already available. Where LLMs have the greatest potential to steer you wrong is when you ask something where the answer is not obvious, the question might be ill-formed, or the user is incorrectly convinced that something should be possible (or easy) when it isn't. Such cases have a lot more in common with these "nonsensical riddles" than they do with any possible frontier benchmark.

This is especially obvious when viewing the reasoning trace for models like Claude, which often spends a lot of time speculating about the user's "hints" and trying to parse out the intent of the user in asking the question. Essentially, the model I use for LLMs these days is to treat them as very good "test takers" which have limited open book access to a large swathe of the internet. They are trying to ace the test by any means necessary and love to take shortcuts to get there that don't require actual "reasoning" (which burns tokens and increases the context window, decreasing accuracy overall). For example, when asked to read a full paper, focusing on the implications for some particular problem, Claude agents will try to cheat by skimming until they get to a section that feels relevant, then searching directly for some words they read in that section. They will do this even if told explicitly that they must read the whole paper. I assume this is because the vast majority of the time, for the kinds of questions that they are trained on, this sort of behavior maximizes their reward function (though I'm sure I'm getting lots of details wrong about the way frontier models are trained, I find it very unlikely that the kinds of prompts that these agents get very closely resemble data found in the wild on the internet pre-LLMs).

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#793

Post by nearbuy »

hansmayer wrote: post #328
nearbuy wrote: post #235

Earlier quoted context omitted.

That's not what they said. They said: > It's evaluation function simply returned the word "Most" as being the most likely first word in similar sentences it was trained on. Which is false under any reasonable interpretation. They do not just return the word most similar to what they would find in their training data. They apply reasoning and can choose words that are totally unlike anything in their training data. If…

You are actually proving my point with your example, if you think about it a bit more.

If there is no response it could give that will disprove your point, then your belief is unfalsifiable and your point is meaningless.

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#794

Post by vlovich123 »

lamonade wrote: post #570
x3ro wrote: post #510

Earlier quoted context omitted.

It's a trick question, humans use these all the time. E.g. "A plane crashes right on the border between Austria and Switzerland. Where do you bury the survivors?" This is not dishonest, it just tests a specific skill.

What’s funny is that it can answer that correctly, but it fails on ”A plane crashes right on the border between Austria and Switzerland. Where do you bury the dead?”

For me when I asked this (but with respect to the border between Austria and Spain) Claude still thought I was asking the survivors riddle and ChatGPT thought I was asking about the logistics. Only Gemini caught the impossibility since there’s no shared border.

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#795

Post by conductr »

WarmWash wrote: post #630
conductr wrote: post #202

Earlier quoted context omitted.

> that is implicitly so easy to grasp for most humans I feel like this is the trap. You’re trying to compare it to a human. Everyone seems to want to do that. But it’s quite simple to see LLMs are quite far still from being human. The can be convincing at the surface level but there’s a ton of nuance that just shouldn’t be expected. It’s a tool that’s been tuned and with that tuning some models will do better than ot…

>But it’s quite simple to see LLMs are quite far still from being human. At this point I think it's a fair bet that whatever supersedes humans in intelligence, likely will not be human like. I think that their is this baked-in assumption that AGI only comes in human flavor, which I believe is almost certainly not the case. To make an loose analogy, a bird looks at a drone an scoffs at it's inability to fly quietly or…

> I believe is almost certainly not the case.

Agree. It's Altman's "Quiet Dominance / Over-reliance / Silent Surrender" risks [0]. Feel this is extremely likely and has already happened to some degree with technology in general and AI will be more pervasive in allowing people to vibe their life decisions, likely with unintended consequences. Vibe coding works because it's quick to change/edit/throw away, but that doesn't generalize well to the real and physical world.

Also should point out this is acceptable because it's just a contrived example of bad LLM-fu. Just like you wouldn't search Google for closest carwash and ask if you should take your car if you knew the answers already. Instead, you'd ask if it's open, does it do full details, what are the prices, etc. Many people with bad Google-fu have problems finding answers to their questions too and that's continued for the past couple decades of it's dominance for information seeking.

[0] Altman describes a more subtle, long-term threat where AI becomes deeply integrated into societal, political, and economic decision-making. He worries that society will become overly dependent on AI, trusting its reasoning over human judgment, leading to a "silent surrender" of human agency.

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#796

Post by thinkling »

jader201 wrote: post #679
FatherOfCurses wrote: post #666

All the people responding saying "You would never ask a human a question like this" - this question is obviously an extreme example. People regularly ask questions that are structured poorly or have a lot of ambiguity. The point of the poster is that we should expect that all LLM's parse the question correctly and respond with "You need to drive your car to the car wash." People are putting trust in LLM's to provide…

That’s not the problem with this post. The problem is that most LLM models answer it correctly (see the many other comments in this thread reporting this). OP cherry picked the few that answered it incorrectly, not mentioning any that got it right, implying that 100% of them got it wrong.

You can see up-thread that the same model will produce different answers for different people or even from run to run.

That seems problematic for a very basic question.

Yes, models can be harnessed with structures that run queries 100x and take the "best" answer, and we can claim that if the best answer gets it right, models therefore "can solve" the problem. But for practical end-user AI use, high error rates are a problem and greatly undermine confidence.

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#797

Post by lukeasch21 »

vbezhenar wrote: post #752

It makes no sense to walk. So the whole question makes no sense as there's no real choice. It seems that LLM assumes "good faith" from the user side and tries to model the situation where that question actually makes sense, producing answer from that situation. I think that's a valid problem with LLMs. They should recognize nonsense questions and answer "wut?".

That's one of the biggest shortcomings of AI, they can't suss out when the entire premise of a prompt is inherently problematic or unusual. Guardrails are a band-aid fix as evidenced by the proliferation in jailbreaks. I think this is just fundamental to the technology. Grandma would never tell you her dying wish was that you learned how to assemble a bomb.

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#798

Post by MadxX79 »

kingstnap wrote: post #725
smallstepforman wrote: post #676

Earlier quoted context omitted.

As people pointed out, change the distance to 43 meters and you’ll get the walk answer.

It would be interesting to see the answer parametrically change. An equally strange trip question is to say the car wash is 0m, 1m, -10m, 1000000m, orange m, etc.

What I really want is to be able to search through the training dataset to see the n closest hits (cosine distance or something). I think the illusion would very quickly be dispelled that way.

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#799

Post by nearbuy »

fauigerzigerk wrote: post #302
grey-area wrote: post #283

Earlier quoted context omitted.

It’s just not intelligent or reasoning, and this sort of question exposes that more clearly. Surely anyone who has used these tools is familiar with the sometimes insane things they try to do (deleting tests, incorrect code, changing the wrong files etc etc). They get amazingly far by predicting the most likely response and having a large corpus but it has become very clear that this approach has significant limitati…

I agree completely. I'm tempted to call it a clear falsification of any "reasoning" claim that some of these models have in their name. But I think it's possible that there is an early cost optimisation step that prevents a short and seemingly simple question even getting passed through to the system's reasoning machinery. However, I haven't read anything on current model architectures suggesting that their so called…

If you ask a bunch of people the same question in a context where they aren't expecting a trick question, some of them will say walk. LLMs sometimes say walk, and sometimes say drive. Maybe LLMs fall for these kinds of tricks more often than humans; I haven't seen any study try to measure this. But saying it's proof they can't reason is a double standard.

Re: I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

#800

Post by cush »

prasoonds wrote: post #46

I've used LLMs enough that I have a good sense of their _edges_ of intelligence. I had assumed that reasoning models should easily be able to answer this correctly. And indeed, Sonnet and Opus 4.5 (medium reasoning) say the following: Sonnet: Drive - you need to bring your car to the car wash to get it washed! Opus: You'll need to drive — you have to bring the car to the car wash to get it washed! Gemini 3 Pro (mediu…

Chat GPT 5.2 also told me to walk, but only because it made the assumption that because I was even asking the question, I must have multiple cars, and one of my cars was already at the car wash being washed. I clarified and said "I only have one car", and it replied that I obviously need to drive.

--- GPT 5.2

Walk.

Driving 50 m just adds friction: starting a cold engine, waiting, maneuvering, then immediately stopping again — all for a distance that’s ~30–40 seconds on foot. Walking is faster door-to-door, less awkward at the wash entrance, and you avoid lining up a wet car twice.

Only reason to drive that distance: you need to bring the car there anyway (e.g., self-serve bay where the car must be present). Otherwise, walking wins.

--- Me

I only have one car

--- GPT 5.2

Then you obviously drive — the car has to be at the car wash to be washed

Post reply on HN