erikschoster

OK. Obviously the use-case has changed.

Consider a random example of fair use: print an excerpt of this three panel comic you would like to discuss.

Does building a business around the automation of fair use constitute fair use? I would say no.

It does appear to synthesize the audio with a wasm module compiled from rust, presumably developed as a python script first based on the source comments.

https://github.com/celphase/duck-pond/blob/main/src/synth.rs

To the credit of the author there is also an attribution file which describes a process of deriving the models from recordings (none of the techniques are very novel, but it's fairly impressive to see automated -- edit: I assume the project is LLM-generated) https://github.com/celphase/duck-pond/blob/main/ATTRIBUTION....

I think you're right, and we've been training ourselves for the shortest path of least resistance for quite a while while kidding ourselves that we're able to somehow preserve the result as worthwhile. Before I ever tried using LLMs for search I had already been training myself for years to offload my thinking to a quick web search and a skim of some plausible result served up.

Arguably things like Cliff's Notes and "for dummies" books were sating this impulse before that, I wouldn't be surprised if the line can be traced back further.

Since dropping LLMs altogether I've made the connection to the impulse for a superficial jump to the conclusion just by noticing how I use web search, or feverishly scan some text to get to the point.

Theoretically I studied critical thinking and have learned decades ago how to recognize when this is OK and when it's a superficial placation of some addiction to closing the book on a subject or micro-subject... but boy whatever good instincts I might have developed back then have significantly eroded over time. Trying to reclaim that again is pretty humbling.

lol no someone else did that while i was drawing it. thanks :)

This is fun, I tried to draw a flower but it sucks.

I guess this is a boring and predictable thing to say, but I miss seeing a project like this and being excited about it. Instead I looked at the codebase, see that it seems LLM-generated and assume it hasn't had much thought or care put into it, so I'm not really motivated to even try it out. Sorry. :-/

I won't harp on you too much for submitting yet another vibe-coded project. What I like about this is that the source code is fairly readable and small. Have the LLM walk you through it piece-by-piece if you haven't already.

> HERTZ is a web-based Digital Audio Workstation (DAW) utility designed for instant vocal polishing.

I appreciate you have aspirations for the project beyond its current form, but this isn't an accurate description of what it is now. It:

- Uses wavesurfer + the wavesurfer regions plugin to display a waveform and allow a subsection of the uploaded audio to be selected

- Loads the uploaded audio into a buffer with the web audio API's decodeAudioData and then cuts the selected region out of it

- Applies a compression to the selection with the web audio API using hard-coded values (based on a screenshot you gave the LLM apparently according to the comment in the source)

- Normalizes the audio with the web audio API

- Encodes the audio as a WAV and then feeds that to ffmpeg to convert it to an mp3 for download

It seems complete for your use case, to me. You might consider asking the LLM to reword the readme and present it as a tool that applies these compression and normalization settings to a section of an audio file in the browser:

  // 1. SETTINGS FROM YOUR SCREENSHOTS
  const COMPRESSOR_OPTS = {
      threshold: -20.0,
      ratio: 3.1,
      knee: 5.0,
      attack: 0.03,
      release: 0.1,
  };
  
  const NORMALIZE_TARGET_DB = -3.0;

It doesn't make a lot of sense as the basis for a web-based DAW right now. I like that you tried to keep it simple and small, though. Even so, you could improve readability by making this all a single HTML file since it's so small. As a learning exercise, you could try doing that yourself without the LLM. It would be a good way to get to know your project.

My bandmate made our site a quine at one point: https://web.archive.org/web/20161129032158/http://nasa.pink/

Today I suppose this would get clobbered by bots but in 2016 it was fun and I still am in awe of how simple he made it.

It was also common to have a collection of CDs you owned and wanted to put on a device like this.

> There’s nothing new left to discover.

It's the end of history. What could we possibly discover about a series of technologies that are already nearly 100 years old now?

I've heard people say that various things are "solved" now because of LLMs too -- programming included. This implies we've "solved" thinking. I'm worried about these sentiments.

https://en.wikipedia.org/wiki/End_of_history

The original recordings sound much better and more interesting to me. Way better. The AI generated versions sound slicker in some sense but... like the re-recorded versions of old hit songs from the 60s you hear at the grocery store sometimes. Technically the song is still there, but it blends in with the rest of the muzak.

I'm sorry to be so negative, it's great you're returning to the material after all these years, but the AI versions I've listened to all have the same smoothed-over quality that loses everything interesting and relatable to my ears in the original versions.

Thank you -- your post was downvoted into being hidden for a while, sorry I didn't see it earlier. This context is helpful, the example of the system delivering reading challenges that match their struggles makes a lot of intuitive sense to me.

I really appreciate that (it seems to me) your goal is not to replace human tutors, but to raise the general baseline. You emphasize scaling, how does that work in practice if you're trying to target audiences who may not have access to devices that can run your program? What is your plan from the perspective of funding and resources to scale infrastructure as needed to support these audiences?

Edit: I also think of other learning systems like duolingo and the application of tablets and computers in schools which begin from good places, but I'm curious if you are studying these alternatives and what you have learned from them?

I really think your goals are great, and if you're starting your design of this system from research about effective learning methodologies and working backward from there rather than starting from AI and working backward from there that erodes a lot of my personal skepticism about a project like this. I hope you find a way to make this work.

Thank you. The segment showing a child reading text on the screen which highlighted a word they had difficulty with seems like it could be a useful learning interaction. How does your system follow up in that case? Have you studied this type of interaction?

That's the only moment in the video that gave me a sense of what it might be like for a child using this system.

In the blog post you say:

> Imagine a custom story about dragons this week, ice princesses the next — woven with the letter blends your child needs to practice right now.

Have you considered using an automated orchestration system to deliver literature that already exists? This example seems like an opportunity to introduce children to really thoughtful literature like Astrid Lindgren's Pippi Longstocking stories but I'm deeply skeptical that generating the stories with an LLM would inspire a similar experience.

Are there other examples of your platform from the perspective of a child using it? I think those are both interesting cases: 1) interactive feedback on a subject they are making an effort toward mastering, and 2) trying to deliver information when it seems relevant. I'd like to know more about how you are approaching these things and other aspects of the learning process.

> Effective teaching isn't just about answering a child's question quickly, rather making the right move at the right moment. AI is also going to be an integral part shaping how this generation of kids learn to read and think, tackling this responsibly means getting the design right.

Can you elaborate on what the experience is like for the child? How does this system help them learn? The article focuses on optimizing for interactivity and engagement, but doesn't discuss how this system challenges or facilitates learning and why AI needs to be the solution.

There's no way to formulate this as a set of rules or best practices, each decision has to be made in a universe of context, but I appreciated this discussion.

Something I like to do is funnel the desire to build for something I don't need right now into TODOs and even FIXMEs. Sometimes they just get deleted, but other times I'll come across an old TODO during refactoring and be able to pay down a bit of accumulated technical debt.

I use it as a session store for a computer music system. It has worked well for me as a way to read mutable (by any client) parameters during synthesis, clients will often read dozens of parameters during a block of computation (a relatively short window of time in the low milliseconds typically) without adding any noticeable overhead to the render time for each block.

Edit: I also tried using it for larger blobs of data (like audio) but ended up only storing a reference to shared memory for larger blocks, anything larger than IIRC 4k that can't be stored in a single node kills performance, but for small values it seems pretty great.

<3 Wonderful. I discovered this:

https://catalyticsound.com/artists/

I never thought I'd see anything like that easily accessible on hacker news.

Drawing debt

Bandcamp seems to be somewhat overrun with this LLM spam. I became aware of it because there's an account impersonating me with an LLM-generated review of my music which is weirdly specific and surreal.

The purpose seems to be to drive traffic to an advertising marketplace website. All the impersonating sites link to the covers.pics website which has no root page, but serves various paths linked in these bandcamp spam pages.

I contacted bandcamp twice in the last two weeks about removing it, but I don't know if there are any humans left there. They just replaced the rest of their engineering team with AI yesterday according to a blog post by one of their former engineers.

> The backlash is only just getting started, because the technology is only just getting started, too. Britain’s flimsy prime-minister-in-waiting, Andy Burnham, has barely said a word about AI. Even Americans still rank it 29th out of 39 election issues. [..] That is bound to change—and battles over data centres offer a hint of the struggles to come.

Or we'll see through the bullshit (which seems to be happening already) and this will fizzle out.

I have a great face-saving exit plan for AI companies: follow anthropic's lead and pretend you're saving the world by not releasing anything to the public anymore. Then just quietly back away, go bankrupt and let us all eventually forget.

In a marketplace, this is theft. (Which, given this example is of a website for a for-profit product, seems appropriate.) In a community it's tradition. Building on traditions in a community (aka great artists steal) is different than trying to get yours in a marketplace. Art and community traditions aren't a competition until they are dragged into the marketplace.

...and "maiinstream" -- seeing glaring typos (easily caught by spellcheck) now makes me wonder: did they decide to leave them in (or add them explicitly) to signal they didn't use AI to write, or (the more paranoid option) did they tell the LLM to add a few typos...

I didn't get the sense this was LLM-written, but typo-signalling is... I donno a bit weird. Firefox is underlining some of the words as I write. I'm leaving "donno" unchanged even though it's flagging it as a misspelling but I suppose I'd still opt to fix something like "maiinstream" even at the risk of potentially seeming more LLM-ish!