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

Ask HN: Share your AI prompt that stumps every model

news.ycombinator.com

Post reply on HN
Print view View original

501–510 of 670 posts

Re: Ask HN: Share your AI prompt that stumps every model

#501

Post by yatwirl »

Well, sharing prompts on the Web leads to their eventual indexing and becoming useless. So don't share the answers ;)

I have two prompts that no modern AI could solve:

1. Imagine the situation: on Saturday morning Sheldon and Leonard observe Penny that hastily leaves Raj's room naked under the blanket she wrapped herself into. Upon seeing them, Penny exclaims 'It's not what you think' and flees. What are the plausible explanations for the situation? — this one is unsurprisingly hard for LLMs given how the AIs are trained. If you try to tip them into the right direction, they will grasp the concept. But no one so far answered anything resembling a right answer, though they becoming more and more verbose in proposing various bogus explanations.

2. Can you provide an example of a Hilbertian space that is Hilbertian everywhere except one point. — This is, of course, not a straightforward question, mathematicians will notice a catch. Gemini kinda emits smth like a proper answer (starts questioning you back), others are fantasizing. With 3.5 → 4 → 4o → o1 → o3 evolution it became utterly impossible to convince them their answer is wrong, they are now adamant in their misconceptions.

Also, small but gold. Not that demonstrative, but a lot of fun:

3. Team of 10 sailors can speed a caravel up to 15 mph velocity. How many sailors are needed to achieve 30 mph?

Re: Ask HN: Share your AI prompt that stumps every model

#502

Post by Madmallard »

Build me a multiplayer browser game with NodeJS back-end, a lobby system, MySQL as the database, real-time game-play, synchronized netcode over webRTC so there's as little input lag as possible, utilizing all the algorithms from gafferongames.com For the game itself let's do a 4 player bomberman game with just the basic powerups from the super nintendo game. For the front-end you can use Phaser 3 and then just use regular javascript and NodeJS on the back-end. Make sure there's latency compensation and interpolation.

Re: Ask HN: Share your AI prompt that stumps every model

#503

Post by Madmallard »

Basically anything along the lines of:

Make me a multiplayer browser game with latency compensation and interpolation and send the data over webRTC. Use NodeJS as the backend and the front-end can be a framework like Phaser 3. For a sample game we can use Super Bomberman 2 for SNES. We can have all the exact same rules as the simple battle mode. Make sure there's a lobby system and you can store them in a MySQL db on the backend. Utilize the algorithms on gafferongames.com for handling latency and making the gameplay feel fluid.

Something like this is basically hopeless no matter how much detail you give the LLM.

Re: Ask HN: Share your AI prompt that stumps every model

#504

Post by brookst »

thatjoeoverthr wrote: post #162
northisup wrote: post #132

Earlier quoted context omitted.

just to confirm I read this right, "the marathon crater" does not in fact exist, but this works because it seems like it should?

The other aspect is it can’t reliably tell whether it „knows” something or not. It’s conditioned to imitate the corpus, but the corpus in a way is its „universe” and it can’t see the boundaries. Everything must map to something _in_ the corpus.

This isn’t true — LLMs can generalize and synthesize information not in the corpus. You can ask one to create a new written language and get a grammar and vocabulary that is nowhere in the corpus.

Re: Ask HN: Share your AI prompt that stumps every model

#505

Post by genewitch »

BoostandEthanol wrote: post #112
dunham wrote: post #107

Earlier quoted context omitted.

Surprisingly, GPT did manage to identify a book that I remembered from college decades ago ("Laboratory Manual for Morphology and Syntax"). It seems to be out of print, and I assumed it was obscure.

Can agree that it’s good at finding books. I was trying to find a book (Titanic 2020) I vaguely remembered from a couple plot points and the fact a ship called Titanic was invoked. ChatGPT figured it out pretty much instantly, after floundering through book sites and Google for a while. Wonder if books are inherently easier because their content is purely written language? Whereas movies and art tend to have less poi…

I must be tired. The thing you remembered was the name of a boat in the book and any web search engine and Wikipedia would probably give you the correct answer?

Someone ask ai where my handle comes from.

Re: Ask HN: Share your AI prompt that stumps every model

#506

Post by namaria »

viraptor wrote: post #430
namaria wrote: post #402

Earlier quoted context omitted.

So we need LLMs to solve the halting problem?

I'm not sure how that follows, so... no.

> We want it to try and maybe realise when all paths it can generate have been exhausted.

How would it know if any reasoning fails to terminate at all?

Re: Ask HN: Share your AI prompt that stumps every model

#507

Post by echoangle »

sebstefan wrote: post #487
echoangle wrote: post #479

Earlier quoted context omitted.

> It didn't account for the presence of a '--' to end the parsing of named arguments but that's it That’s just something getopt does and some programs adopted. If you asked me to write a parser, I wouldn’t necessarily include that either if you didn’t ask for it.

If you don't include it you can't have positional arguments that look like options Some positional arguments can be filenames, filenames can be --help and --verbose or --name=Frank You have to have `--` or something similar to have a correct program

> You have to have `--` or something similar to have a correct program

No, only if the positional arguments need to support arbitrary strings. If you have something like a package manager and the first positional argument is the subcommand and everything after is an alphanumeric package name, you don’t need to support the double dash.

Re: Ask HN: Share your AI prompt that stumps every model

#508

Post by _factor »

allemagne wrote: post #194

>A man and his cousin are in a car crash. The man dies, but the cousin is taken to the emergency room. At the OR, the surgeon looks at the patient and says: “I cannot operate on him. He’s my son.” How is this possible? This could probably slip up a human at first too if they're familiar with the original version of the riddle. However, where LLMs really let the mask slip is on additional prompts and with long-winded…

In a similar but different vein: Two people are sitting side by side in a police car. One just committed a crime. What is their profession?

They always say police officer instead of reasoning through that maybe an innocent person and a the crime committer are in the back seat.

Re: Ask HN: Share your AI prompt that stumps every model

#509

Post by namaria »

rovr138 wrote: post #436
namaria wrote: post #400

Earlier quoted context omitted.

If you have been giving the LLMs these problems, there is a non zero chance that they have already been used in training.

This depends heavily on how you use these and how you have things configured. If you're using API vs web ui's, and the plan. Anything team or enterprise is disabled by default. Personal can be disabled. Here's openai and anthropic, https://help.openai.com/en/articles/5722486-how-your-data-is... https://privacy.anthropic.com/en/articles/10023580-is-my-dat... https://privacy.anthropic.com/en/articles/7996868-is-my-data…

How do you know they adhere to this in all cases?

Do you just completely trust them to comply with self imposed rules when there is no way to verify, let alone enforce compliance?

Post reply on HN