Hacker News Reader

rrherron TuneNN: A transformer-based network model for pitch detection

How does the accuracy of this compare to CREPE?

https://github.com/marl/crepe

https://github.com/maxrmorrison/torchcrepe

Does anyone know what the current state of the art is, within the Music Information Retrieval community?

rrherron OpenAI Investors Scramble to Reinstate Sam Altman as CEO

Kind of, but the Forbes post says more about Microsoft than The Verge post

rrherron OpenAI's board has fired Sam Altman

That you, ChatGPT? You're hired!

rrherron Ship Shape

"We developed a variation on the Ramer-Douglas-Peucker (RDP) algorithm, which is a curve simplification algorithm that reduces the number of points in a curve while preserving its important details. It achieves this by recursively removing points that deviate insignificantly from the simplified version of the curve."

This reminded me of an old side project, which others may be interested in. I applied Douglas-Peucker to Picasso for a talk at Strange Loop 2018:

Picasso's Bulls: Deconstructing his design process with Python https://rrherr.github.io/picasso/

rrherron Louis Armstrong's Last Word

The full quote in context was: "You can name a thousand great instrumentalists or you can name a thousand great vocalists, but he’s the only person you could find who changed the way people played their instruments and the way people sang. Louis does that in a four-year period in the 1920s; by 1930, if you aren’t playing or singing like him, you’re out of work."

rrherron Ask HN: What's the book that people in your field pretend to have read?

The Elements of Statistical Learning

rrherron What Ghosting Says About Society (and Why It Hurts So Much)

See also:

"A shift in American family values is fueling estrangement"

https://news.ycombinator.com/item?id=27833159

rrherron Show HN: Improved freemusicdemixer – AI music demixing in the browser

Oh yes, that would be amazing! Good luck!

rrherron Fine-tune your own Llama 2 to replace GPT-3.5/4

“most tasks can be formulated this way, including autocomplete tasks”

For autocomplete tasks, with a corpus of unlabeled documents, would you insert a separator token at an arbitrary space in each document, in order to form input/output pairs?

rrherron Fine-tune your own Llama 2 to replace GPT-3.5/4

"You do this by training an existing model on example input/output pairs that demonstrate the task you want your fine-tuned model to learn."

Are fine-tuning datasets required to be input/output pairs? Or instead, can the fine-tuning be autoregressive (predict the next token throughout this corpus of unlabeled documents)?

rrherron A study of malicious code in PyPI ecosystem

Agreed about Docker.

Personally, I use Google Colab for my convenient, lightweight Python sandbox. But I’d rather not depend on Google. It feels weird that this is my best option for now.

rrherron Become an expert in something specific and boring

Peter Singer, nope.

https://www.nytimes.com/2003/02/16/magazine/unspeakable-conv...

https://archive.ph/OXXQb

rrherron Shining a light on the digital dark age

https://web.archive.org/web/20230901181858/https://longnow.o...

rrherron Noiszy: A browser plugin that creates meaningless web data – digital “noise.”

in a world where big data threatens to commodify our lives,. telling online surveys that i "Dont know" what pringles are constitutes Heroism

https://twitter.com/dril/status/1061091401251745792

rrherron Llama 2 on ONNX runs locally

How does this compare to using https://github.com/ggerganov/llama.cpp with https://huggingface.co/models?search=thebloke/llama-2-ggml ?

rrherron The Four Hobbies, and Apparent Expertise

The 2nd diagram, under the heading "Appearance of Expertise", shows a negative correlation between expertise doing the hobby versus discussing the gear.

Counterpoint: "Jean Renoir once told us that when his father's friends used to come to the house — all the distinguished French painters of the time — they never talked about grand theories of painting; they talked about where to get the best turpentine"

https://quoteinvestigator.com/2016/10/20/turpentine/

rrherron Dril is everyone. More specifically, he’s a guy named Paul

This is dril's first interview under his real name and not "in character."

rrherron Why human societies developed so little during 300k years

Yes, The Gossip Trap is a great read!

Here's an excerpt from the intro, for context:

"This essay of mine won 1st place and $5,000 in Scott Alexander’s anonymous book review contest, which was hosted at Astral Codex Ten.

It attempts to solve the “Sapient Paradox,” which asks: Given that humans have been around for 200,000 years, why did civilization take so long to get started? Why were we stuck in prehistory for so long?"

rrherron User interface design timeframes: from 0.1 seconds to 10 years (2009)

See also “10 Timeframes” by Paul Ford:

https://contentsmagazine.com/articles/10-timeframes/

This is Paul Ford’s insightful and poignant keynote talk for the 2012 graduating class of the Interaction Design MFA program at the School of Visual Arts.

I’ve been haunted by these lines:

“I can never remember if we are supposed to live each day as it were our last, or if it’s the first day of the rest of our lives. It’s hard to tell sometimes. We make movies about it over and over again. … Of course these movies are made by people who are totally dedicated to making films. They give up their lives and neglect their children to make movies about the value of family.”

rrherron Is there a market failure in child care?

"Let us assume that the parents get neither utility nor disutility from spending time with their infants."

Umm...

rrherron MusicLM: Generating music from text

> Anyone want to share their experiments with MusicLM, feel free to join

How can we experiment with MusicLM?

The paper’s last sentence says, “We have no plans to release models at this point.”

rrherron Stripe laying off around 14% of workforce

Then for your education section, do you keep the dates there, or remove those dates?

rrherron Obsidian Dataview: Turn Obsidian Vault into a database you can query from

You can sort Dataview results using this plugin: https://github.com/alexandru-dinu/obsidian-sortable

rrherron No one cared about my spreadsheets

> If you are an employer and see a student that dropped out of college without a very good reason, then that is a very strong negative signal.

This is an argument for Caplan’s signaling theory, not against it. You literally said that whether or not a job applicant graduated is a very strong signal to employers.

rrherron Putting Ideas into Words

Jesus is another example

rrherron Ask HN: Anybody else exhausted by code discussion?

Sounds like you're more interested in "solving the business problem" and less interested in "solving the problems of the code that is solving the business problem."

Hillel Wayne recently wrote about this distinction:

> New Newsletter! An exploration of a core difference between "programming" and "software engineering": how much you work on "solving the business problem" versus "solving the problems of the code that is solving the business problem."

https://twitter.com/hillelogram/status/1486079674795143170

https://buttondown.email/hillelwayne/archive/software-artifa...

And sounds like you enjoy "product engineering" more than "infrastructure engineering."

Michelle Lim wrote a post about it, which was discussed on HN.

> Try using the "Product/Infrastructure" axis as the first axis to understand your career preference.

> Broadly speaking, there are 2 types of engineers: "Product-first" engineers are obsessed with using code to solve a user problem and they see code as just a means to an end. "Code-first" engineers are obsessed with the abstractions, architecture, tools and libraries in the code. Elegant code is the end.

https://news.ycombinator.com/item?id=28476538

https://www.michellelim.org/writing/stop-using-frontend-back...

rrherron You Don't Need a Mentor–Find a Nemesis Instead

> I will always prefer watching Magic-versus-Bird over those Michael Jordan highlight films—even though I realize the sheer superiority in athleticism of the latter. Without a nemesis, Jordan could win championships, but he couldn’t create the kind of Greek tragic drama of the encounter with the other.

Michael Jordan did have a nemesis 1988-1991: Isiah Thomas.

https://en.wikipedia.org/wiki/Bulls–Pistons_rivalry

rrherron Ask HN: How can I get a tech job that's more social?

Can anyone recommend some remote employers that do pair programming?

rrherron Show HN: Composing Studio – An online, collaborative editor for music notation

Awesome! I have an old project that generates ABC music notation. I made a context-free grammar to transform plain melodies into bluegrass banjo style. The output can be played & edited in Composing Studio: https://rrherr.github.io/banjo-grammar/

For example:

    T:Happy Birthday, bluegrass banjo style
    %%MIDI program 105
    K: G
    L: 1/16
    Q: 140
    D2 D2 E4 D4 G4 F8 D2 D2 E4 D4 A4 G8 D2 D2 d4 B4 G4 F4 E4 c2 c2 B4 G4 A4 G8
    Dd Dd Edgd Ddgd Gdgd FdgFdgFd Dd Dd Edgd Ddgd Adgd GdgGdgGd 
    Dd Dd ddgd Bdgd Gdgd Fdgd Edgd cd cd Bdgd Gdgd Adgd GdgGdgGd

rrherron A plain text note-taking system for problem solving

Can you demonstrate how to use Calculist to solve a specific problem?

I tried the demo and expanded the items in the sample document. I saw examples of calculation! So I’d love to see examples of problem solving with Calculist.