Hacker News Reader

Comment by silenced_trope | Hacker News Reader

I'm the opposite. I feel like it is true. I very much will NO LONGER THINK about "simple" problems.

My most recent thing is for a game I'm working on. I got a bunch of free sound FX from freesound.org, but they were AIFF format. I googled and found an online converter, then asked ChatGPT to use ffmpeg to write a script that goes through the directory converting the AIFFs to MP3s.

In the past if I would have written such a script it would have required me to "look into how to do it" and then I would have saved the script somewhere on my machine or online. Now I just deleted it right away. If I ever need to do it again or something similar I'll just ask ChatGPT and get it from scratch again.

View on HN

> use ffmpeg

Well there's your problem. Such a powerful tool, but that CLI is something else.

Every time I've tried to get it to do something reasonably simple like this, it returns plausible-sounding code that doesn't work.

My last attempt to use it was to do with resolving circular calculations in Excel - I wanted some way of calculating the sum of a geometric series without having the series in actual cells. It kept recommending things that didn't work.

I'm always amazed to hear of people who manage to get working code from an AI. This has not been my experience.

Work on your prompts. Break down the problem into pseudocode steps, the way you might write down comments for yourself on how to do something. Provide examples of similar working code if you can. Don't ask it to do everything at the same time, break it up into limited goal increments.

In my experience LLMs work best if you already know how to do something, but you just can't be bothered to actually do it.

When exploring unknown territory, keep your questions simple and ask for clarification on anything you don't understand. It will not just help you, but the AI as well.

> I wanted some way of calculating the sum of a geometric series without having the series in actual cells.

There are simple explicit formulas for this.

yeah, sumproduct on google docs was the thing it kept suggesting, but not in a way that worked for the thing I wanted. I found a way around it in the end, but I feel I wasted 30 mins trying to get ChatGPT to understand what I wanted and give me something that actually worked.

\sum_{i=0}^n x^i (the sum of all terms x^i as i ranges from 0 to n, inclusive) = (x^{n+1} - 1)/(x-1).

Or, if n is positive infinity, you assume that |x| < 1 and thus the sum is (0 - 1) / (x-1) = 1/(1-x).

If you need the sum from x^m to x^n, that's equal to the sum from 0 to n minus the sum from 0 to (m-1).

You can derive the formula yourself without difficulty:

         sum = 1 + x + x^2 + ... + x^n
    (sum)(x) =     x + x^2 + ... + x^n + x^{n+1}

    (sum)(x) - sum =  x^{n+1} - 1
    (sum)(x-1)     =  x^{n+1} - 1
    (sum)          = (x^{n+1} - 1)/(x-1)

a whole script? surely it's just something like `find . -name '*.aiff' | parallel ffmpeg -i {} {}.mp3`

Doesn’t VLC or Foobar have a batch convert feature? Or Audacity? One of them does, pretty sure?

Reply on news.ycombinator.com