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

“Car Wash” test with 53 models

opper.ai

Post reply on HN
Print view View original

81–90 of 469 posts

Re: “Car Wash” test with 53 models

#81

Post by Terr_ »

ben_w wrote: post #77
estimator7292 wrote: post #36

Earlier quoted context omitted.

I've always wondered about that. LLM providers could easily decimate the cost of inference if they got the models to just stop emitting so much hot air. I don't understand why OpenAI wants to pay 3x the cost to generate a response when two thirds of those tokens are meaningless noise.

Because they don't yet know how to "just stop emitting so much hot air" without also removing their ability to do anything like "thinking" (or whatever you want to call the transcript mode), which is hard because knowing which tokens are hot air is the hard problem itself. They basically only started doing this because someone noticed you got better performance from the early models by straight up writing "think step…

IMO it supports the framing that it's all just a "make document longer" problem, where our human brains are primed for a kind of illusion, where we perceive/infer a mind because, traditionally, that's been the only thing that makes such fitting language.

Re: “Car Wash” test with 53 models

#82

Post by glitchc »

stetrain wrote: post #22
glitchc wrote: post #18

The question does not specify what kind of car it is. Technically speaking, a toy car (Hot wheels or a scaled model) could be walked to a car wash. Now why anyone would wash a toy car at a car wash is beyond comprehension, but the LLM is not there to judge the user's motives.

I think if surveyed at least 90% of native English speakers would understand "I want to wash my car" to mean a full size automobile. The next largest group would probably ask a clarifying question, rather than assume a toy car.

Yes, but you're speaking to a computer, not a person. It, of course, runs into the same limitations that every computer system runs into. In this case, it's undefined/inconsistent behavior when inputs are ambiguous.

Re: “Car Wash” test with 53 models

#84

Post by felix089 »

gs17 wrote: post #79
sambaumann wrote: post #38

Earlier quoted context omitted.

I feel like this has gotten much worse since they were introduced. I guess they're optimizing for verbosity in training so they can charge for more tokens. It makes chat interfaces much harder to use IMO. I tried using a custom instruction in chatGPT to make responses shorter but I found the output was often nonsensical when I did this

Yeah, ChatGPT has gotten so much worse about this since the GPT-5 models came out. If I mention something once, it will repeatedly come back to it every single message after regardless of if the topic changed, and asking it to stop mentioning that specific thing works, except it finds a new obsession. We also get the follow up "if you'd like, I can also..." which is almost always either obvious or useless. I occasion…

It's similar for me, it generates so much content without me asking. if I just ask for feedback or proofreading smth it just tends to regenerate it in another style. Anything is barely good to go, there's always something it wants to add

Re: “Car Wash” test with 53 models

#85

Post by hmokiguess »

streetfighter64 wrote: post #80
hmokiguess wrote: post #75

Earlier quoted context omitted.

No. I would expect it to say 4 given that has an objective answer. For the other, without any context whatsoever, I would prefer the answer of clarifying. I would be okay if the way it asked for clarification came with: “What do you mean walk or drive? I don’t understand the options given you would need your car at the car wash. Is there something else I should know?”

"What do you mean two plus two? I don't understand the question given that it's basic math. Is there something else I should know?"

I fail to see how these things are one and the same. I get the point you are making, I just don't agree with it.

2+2 is a complete expression, the other is grammatically correct but logically flawed. Where is the logical fallacy in 2+2?

Re: “Car Wash” test with 53 models

#86

Post by cadamsdotcom »

This is a beautiful example of a little prompt engineering going a long way

I asked Gemini and it got it wrong, then on a fresh chat I asked it again but this time asked it to use symbolic reasoning to decide.

And it got it!

The same applies to asking models to solve problems by scripting or writing code. Models won’t use techniques they know about unprompted - even when it’ll result in far better outcomes. Current models don’t realise when these methods are appropriate, you still have to guide them.

Re: “Car Wash” test with 53 models

#87

Post by kayge »

layer8 wrote: post #14

> This is a trivial question. There's one correct answer and the reasoning to get there takes one step: the car needs to be at the car wash, so you drive. I don’t think it’s that easy. An intelligent mind will wonder why the question is being asked, whether they misunderstood the question, or whether the asker misspoke, or some other missing context. So the correct answer is neither “walk” nor “drive”, but “Wat?” or…

Yep, just a little more context and all/most of the models would do much better. And sure, most average+ intelligence adults whose first language is English (probably) don't need this, but they're not the target audience for the instructions :)

"The 'car wash' is a building I need to drive through."

or

"The 'car wash' is a bottle of cleaning fluid that I left at the end of my driveway."

https://i5.walmartimages.com/seo/Rain-x-Foaming-Car-Wash-Con...

Re: “Car Wash” test with 53 models

#88

Post by ben_w »

Terr_ wrote: post #81
ben_w wrote: post #77

Earlier quoted context omitted.

Because they don't yet know how to "just stop emitting so much hot air" without also removing their ability to do anything like "thinking" (or whatever you want to call the transcript mode), which is hard because knowing which tokens are hot air is the hard problem itself. They basically only started doing this because someone noticed you got better performance from the early models by straight up writing "think step…

IMO it supports the framing that it's all just a "make document longer" problem, where our human brains are primed for a kind of illusion, where we perceive/infer a mind because, traditionally, that's been the only thing that makes such fitting language.

To an extent. Even though they're clearly improving*, they also definitely look better than they actually are.

* this time last year they couldn't write compilable source code for a compiler for a toy language, I know because I tried

Re: “Car Wash” test with 53 models

#89

Post by felix089 »

cadamsdotcom wrote: post #86

This is a beautiful example of a little prompt engineering going a long way I asked Gemini and it got it wrong, then on a fresh chat I asked it again but this time asked it to use symbolic reasoning to decide. And it got it! The same applies to asking models to solve problems by scripting or writing code. Models won’t use techniques they know about unprompted - even when it’ll result in far better outcomes. Current m…

Interesting, which Gemini model? And how did you ask for symbolic reasoning, just added it to the prompt?

Re: “Car Wash” test with 53 models

#90

Post by stetrain »

glitchc wrote: post #82
stetrain wrote: post #22

Earlier quoted context omitted.

I think if surveyed at least 90% of native English speakers would understand "I want to wash my car" to mean a full size automobile. The next largest group would probably ask a clarifying question, rather than assume a toy car.

Yes, but you're speaking to a computer, not a person. It, of course, runs into the same limitations that every computer system runs into. In this case, it's undefined/inconsistent behavior when inputs are ambiguous.

Yes, but part of the value of LLMs is that they are supposed to work by talking to them like a human, not like a computer.

I could already talk to a computer before LLMs, via programming or query languages.

Post reply on HN