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

I'd rather read the prompt

claytonwramsey.com

Post reply on HN
Print view View original

31–40 of 880 posts

Re: I'd rather read the prompt

#31

Post by vunderba »

bost-ty wrote: post #4

I like the author's take: it isn't a value judgement on the individual using ChatGPT (or Gemini or whichever LLM you like this week), it's that the thought that went into making the prompt is, inevitably, more interesting/original/human than the output the LLM generates afterwards. In my experiments with LLMs for writing code, I find that the code is objectively garbage if my prompt is garbage. If I don't know what I…

This is the CRUX of the issue. Even with SOTA models (Sonnet 3.5, etc) - the more open-ended your prompt - the more banal and generic the response. It's GIGO turtles all the way down.

I pointed this out a few weeks ago with respect to why the current state of LLMs will never make great campaign creators in Dungeons and Dragons.

We as humans don't need to be "constrained" - ask any competent writer to sit quietly and come up with a novel story plot and they can just do it.

https://news.ycombinator.com/item?id=43677863

That being said - they can still make AMAZING soundboards.

And if you still need some proof, crank the temperature up to 1.0 and pose the following prompt to ANY LLM:

  Come up with a self-contained single room of a dungeon that involves an 
  unusual puzzle for use with a DND campaign. Be specific in terms of the 
  puzzle, the solution, layout of the dungeon room, etc. It should be totally 
  different from anything that already exists. Be imaginative.
I guarantee 99% of the returns will return a very formulaic physics-based puzzle response like "The Resonant Hourglass", or "The Mirror of Acoustic Symmetry", etc.

Re: I'd rather read the prompt

#32

Post by nathants »

the solution is obvious. stop grading the result, and start grading the process.

if you can one-shot an answer to some problem, the problem is not interesting.

the result is necessary, but not sufficient. how did you get there? how did you iterate? what were the twists and turns? what was the pacing? what was the vibe?

no matter if with encyclopedia, google, or ai, the medium is the message. the medium is you interacting with the tools at your disposal.

record that as a video with obs, and submit it along with the result.

for high stakes environments, add facecam and other information sources.

reviewers are scrubbing through video in an editor. evaluating the journey, not the destination.

Re: I'd rather read the prompt

#33

Post by scarface_74 »

StefanBatory wrote: post #25
scarface_74 wrote: post #10

I don’t know anything about the subject area, so I don’t know if this captures enough to get a good grade. But I’m curious if anyone could tell whether the last answer were AI generated if I copied and pasted. These are the iterations I go through when writing long requirement documents/assessments/statements of work (consulting). Yes I know the subject area for which I write assessments and know if what is generated…

To me, this part > I didn’t realize how much that could throw things off until I saw an example where the object started moving in a strange way when it hit that point. Would feel off, because why change the person? And even if it's intented, then I'd say it's not formal to do in an assignement.

These are art students not English writers. If I were a teacher I would think this is more authentic. LLMs don’t make this kind of mistake in its default house style.

Re: I'd rather read the prompt

#34

Post by tomjen3 »

I wish the author had state out right that they were not using LLMs much, since their opinion on them and their output has no value (its a new technology, and different enough that you do have to spend some time with them in order to be able to find out what value they have for your particluar work[0].

The is especially the case when you are about to complain about style, since that can easily be adjusted, by simply telling the model what you want.

But I think there is a final point that the author is also wrong about, but that is far more interesting: why we write. Personally I write for 3 reasons: to remember, to share and to structure my thoughts.

If an LLM is better then me at writing (and it is) then there is no reason for me to write to communicate - it is not only slower, it is counterproductive.

If the AI is better at wrangling my ideas into some coherent thread, then there is no reason for me to do it. This one I am least convinced about.

AI is already much better than me at strictly remembering, but computers have been that since forever, the issue is mostly convinient input/output. AIs makes this easier thanks to speech to text input.

[0]: See eg. https://www.oneusefulthing.org/p/centaurs-and-cyborgs-on-the....

Re: I'd rather read the prompt

#35

Post by echelon »

Ancalagon wrote: post #8

I fully support the author’s point but it’s hard to argue with the economics and hurdles around obtaining degrees. Most people do view obtaining a degree as just a hurdle to getting a decent job, that’s just the economics of it. And unfortunately the employers these days are encouraging this kind of copy/paste work. Look at how Meta and Google claim the majority of the new code written there is AI created? The world…

> I fully support the author’s point

I don't. I think the world is falling into two camps with these tools and models.

> I now circle back to my main point: I have never seen any form of create generative model output (be that image, text, audio, or video) which I would rather see than the original prompt. The resulting output has less substance than the prompt and lacks any human vision in its creation. The whole point of making creative work is to share one’s own experience

Strong disagree with Clayton's conclusion.

We just made this with AI, and I'm pretty sure you don't want to see the raw inputs unless you're a creator:

https://www.youtube.com/watch?v=H4NFXGMuwpY

I think the world will be segregated into two types of AI user:

- Those that use the AI as a complete end-to-end tool

- Those that leverage the AI as tool for their own creativity and workflows, that use it to enhance the work they already do

The latter is absolutely a great use case for AI.

Re: I'd rather read the prompt

#36

Post by matthewdgreen »

robertlagrant wrote: post #18

Time to go back to writing essays in exams, live, on paper.

It's challenging. Assignments (and particularly programming assignments) were by far the larger and more difficult part of my CS degree, and also the place where I learned the most. I cannot imagine losing that portion of my education and just replacing it with a few exams.

Re: I'd rather read the prompt

#37

Post by oncallthrow »

LLM cheating detection is an interesting case of the toupee fallacy.

The most obvious ChatGPT cheating, like that mentioned in this article, is pretty easy to detect.

However, a decent cheater will quickly discover ways to conduce their LLM into producing text that is very difficult to detect.

I think if I was in the teaching profession I'd just leave, to be honest. The joy of reviewing student work will inevitably be ruined by this: there is 0 way of telling if the work is real or not, at which point why bother?

Re: I'd rather read the prompt

#38

Post by bonoboTP »

firefoxd wrote: post #29

Personally, I've used LLM to help me better structure my blog post after I write it. Meaning I've already written it, then it enhances it. Most of the time, I'm happy with the results at the time of editing. But when I come back a week or two to re-read it, it looks just like the example the author shared. The goal is to make something legible, but the reality is we are producing slop. I'm back to writing before my b…

[Edit: I agree] I've also grown to dislike even this use case. I did this back in 2023 but as AI text is spreading, the style - yes, even with prompt adjustments it leaks through - is recognized by more and more people and it's a very very bad look. If I see AI-like text from someone, I take it as an insult. It means they don't feel that it's worth their time to brush up the text themselves. And sure, it may well be that they don't value our interaction enough to spend the time on it. But that fact is indeed by itself insulting. So I only send AI touched up text to orgs that are so faceless or bureaucratic that I don't mind "offending" them.

I've grown to respect typos and slightly misconstructed sentences. It's an interesting dynamic that now what appeared lazy to 2021 eyes actually indicates effort and what appeared polished and effortful in 2021 now indicates laziness.

An example is how the admins of my local compute cluster communicate about downtimes and upgrades etc and they are clearly using AI and it's so damn annoying, it feels like biting into cotton candy fluff. Just send the bullet points! I don't need emojis, I don't need the fake politeness. It's no longer polite to be polite. It doesn't signal any effort.

Re: I'd rather read the prompt

#39

Post by mezyt »

mrweasel wrote: post #19
Ancalagon wrote: post #8

I fully support the author’s point but it’s hard to argue with the economics and hurdles around obtaining degrees. Most people do view obtaining a degree as just a hurdle to getting a decent job, that’s just the economics of it. And unfortunately the employers these days are encouraging this kind of copy/paste work. Look at how Meta and Google claim the majority of the new code written there is AI created? The world…

> Most people do view obtaining a degree as just a hurdle to getting a decent job Then fail to actually learn anything and apply for jobs and try to cheat the interviewers using the same AI that helped them graduate. I fear that LLMs have already fostered the first batch of developers who cannot function without it. I don't even mind that you use an LLM for parts of your job, but you need to be able to function witho…

> I fear that LLMs have already fostered the first batch of developers who cannot function without it.

Playing the contrarian here, but I'm from a batch of developers that can't function without a compiler, and I'm at 10% of what I can do without an IDE and static analysis.

Post reply on HN