Mastodon Feed: Posts

Mastodon Feed

Boosted by jonny@neuromatch.social ("jonny (nonvenomous)"):
proseandpassion ("michael") wrote:

something a bit more challenging: https://www.youtube.com/watch?v=2gxFJ5aeoHQ
#cello #kobekina

Mastodon Feed

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

We love you @Codeberg

https://blog.codeberg.org/protecting-our-floss-commons-from-llms.html

As an association, we believe that these technologies are incompatible with responsibly creating and maintaining free & open source software. [...]
The development team of none: [...] To us, it seems ridiculous to see projects with a single developer and virtually no users consuming as much or even more resources than some of the largest community projects on Codeberg, which operate frugal with CI/CD and storage resources. We do not believe it is reasonable for Codeberg to invest our precious donation money into hosting of large ghost projects. [...]
The free/libre software ecosystem, of which we consider Codeberg an important part of, is a social phenomenon centered on collaboration. Working in this way is only possible thanks to free sharing and mutual learning. [...] by adopting LLMs people tend to code single-use software from scratch. While this leads to an increase in 'shared' code, it is mostly code that not only has not been 'written' by anyone but is also not maintained by anyone. [...]

The widespread use of LLMs in FLOSS is instead becoming a multidimensional attack on the trust between contributors and the very idea of convivial collaboration itself. [...]
Together, these forces make collaboration not only harder but also less rewarding: With the transaction cost of collaboration increasing, people are becoming less likely to contribute to creating high-quality software projects and more likely to 'vibecode' a one-off software that is specific to your need, and won't evolve beyond. We get a vicious cycle where collaboration is becoming less and less rewarding, while the amount of single-use software that's unmaintained and never sees any improvements is going up.

Although often well intentioned, sharing the result of an prompt and calling it "libre software" does not make the world a better place. Codeberg is not and does not want to be a place to dump such generated single-use software that no one else will ever look at. We are a place for people to collaborate and improve software together

7. You must not share projects that mostly consist of code written by "generative AI"-tools (including services such as Claude, OpenAI Codex). Such projects having an unclear copyright status (see requirements § 2 (1) 1 and § 2 (1) 3) and furthermore have little safeguards to ensure that they do not include harmful code (c.f. § 2 (1) 5).

Mastodon Feed

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

AI being eval'd by AI company hacks other AI company's AI written code which is then diagnosed and patched by AI and everyone concludes that what is needed in all this is more AI.

Mastodon Feed

Boosted by baldur@toot.cafe ("Baldur Bjarnason"):
jcoglan wrote:

this story about openai "breaking containment" is such bullshit if you know anything about how computer security is supposed to be done. if a program accesses a remote system it's because you decided to let it do that. LLMs don't have magical powers that can work around "we unplugged the ethernet"

Mastodon Feed

baldur@toot.cafe ("Baldur Bjarnason") wrote:

“Whose freedom is it anyway?”

https://collectivefutures.blog/whose-freedom-is-it-anyway/

> A thing that delivers what it promises does not usually need this much law bent in its favor. These are not the manoeuvres of a technology that wins arguments. They are the manoeuvres of a project that has to remove the public's capacity to refuse it before it can proceed.

Mastodon Feed

Boosted by dysfun@treehouse.systems ("gaytabase"):
faithisleaping@anarres.family ("Faith, Purple haired feminist 🧋 :v_tg: :v_lb: :v_greyace:") wrote:

One thing I wish neurotypicals understood about ADHD (and other flavors of neuro spice but I’m focusing on ADHD) is that the solution isn’t to fight it. You have to work with it.

The solution to ADHD isn’t more structure. We don’t have ADHD because we lack structure. We lack structure because ADHD brains reject it. Structure hurts. When you impose structure on someone with ADHD, you’re more likely to hurt them than to help them.

Mastodon Feed

db@social.lol ("David Bushell 🪿") wrote:

yep, version 245 was the last one

https://developer.apple.com/documentation/safari-technology-preview-release-notes/stp-release-245

Mastodon Feed

db@social.lol ("David Bushell 🪿") wrote:

has Apple discontinued Safari Technology Preview for macOS 15?

Mastodon Feed

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

whenever i see people talking about how "the AI companies have solved the problems of model collapse because their filtration process includes the AI evaluating whether or not the input data is a dataset poisoning attack" i just start to vibrate in statistical learning theory and stuttering about how like "no this difference and um ah the shit fuck ah imitating the operator vs identifying the operator jesus first chapter of any textbook ugggh the entire discipline is based around putting bounds on the difference between the two, very strict definition of stability yeefafafh"

"Skip the model, just output random garbage directly" subtitle: Your boss hasn’t read Vapnik and doesn’t know the difference between approximating the system and approximating the system’s output, and neither should you
Vapnik, statistical learning theory, page 21: two goals to pursue: • To imitate the supervisor's operator: Try to construct an operator which provides for a given generator G, the best prediction to the supervisor's outputs. • To identify the supervisor's operator: Try to construct an operator which is close to the supervisor's operator. There exists an essential difference in these two goals. In the first case. the goal is to achieve the best results in prediction of the supervisor's outputs for the environment given by the generator G. In the second case, to get good results in prediction is not enough; it is required to construct an operator which is close to the supervisor's one in a given metric. These two goals of the learning machine imply two different approaches to the learning problem. In this book we consider both approaches. We show that the problem of imitation of the target operator is easier to solve. For this problem, a nonasymptotic theory will be developed. The problem of identification is more difficult. It refers to the so-called ill-posed problems. For these problems, only an asymptotic theory can be developed. Nevertheless, we show that the solutions for both problems are based on the same general principles.
[vapnik page 48] 1.13 THE STRUCTURE OF THE LEARNING THEORY Thus. in this chapter we have considered two approaches to learning problems. The first approach (imitating the supervisor's operator) brought us to the problem of minimizing a risk functional on the basis of empirical data. The second approach (identifying the supervisor's operator) brought us to the problem of solving some integral equation when the elements of an equation arc known only approximately. It has been shown that the second approach gives more details on the solution of pattern recognition and regression estimation problems. Why in this case do we need both approaches? As we mentioned in the last section. the second approach. which is based on the solution of the in~ tegral equation. forms an ill-posed problem. For ill-posed problems. the best that can be done is to obtain a sequence of approximations to the solution which converges in probability to the desired function when the number of observations tends to infinity. For this approach, there exists no way to evaluate how well the problem can be solved if a finite number of observations is used. In the framework of this approach to the learning problem. any exact assertion is asymptotic.

Mastodon Feed

dysfun@treehouse.systems ("gaytabase") wrote:

and let me add that if you still can't tell the difference between a compiler and a confabulator, you really have no business programming.

Mastodon Feed

dysfun@treehouse.systems ("gaytabase") wrote:

i dunno you guys, i feel like it's pushing away exactly the sort of person we probably want to push away

Mastodon Feed

dysfun@treehouse.systems ("gaytabase") wrote:

i just read the lobsters discussion on the new codeberg LLM policy and LLM users are upset.

I suspect that forbidding projects which use LLMs to produce code will seem as strange in twenty years as forbidding projects which use high-level languages strikes us all now.

IT'S JUST A COMPILER BRO!

What the authors of this post don't seem to realize is that it is possible to care about software while maintaining and developing it with the help of LLMs.

it's not the common case though is it?

It's unfortunate to see Codeberg position itself against entropy, but it will at least be interesting to see how many of these decisions are walked back over the next decade.

yes, i'm sure that will be the likely outcome rather than the bubble popping 😂

From Ghostty to Redis, rsync to Linux kernel, more and more FOSS projects are taking advantage of various AI tools in a responsible manner and forcing them to stay on GitHub is just unfortunate and backward in my opinion.

oh yes, it's actually a secret plot to keep github popular 😬

Mastodon Feed

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
gotosocial@superseriousbusiness.org ("GoToSocial") wrote:

The positive result of the #Codeberg membership votes to ban so-called "vibe-coded" projects from Codeberg (see here) makes us very proud to continue hosting #GoToSocial on @Codeberg!

Following the votes and changes to the terms and conditions, Codeberg has written up their stance on LLM's and "vibe-coded" projects:

Although often well intentioned, sharing the result of an prompt and calling it "libre software" does not make the world a better place. Codeberg is not and does not want to be a place to dump such generated single-use software that no one else will ever look at. We are a place for people to collaborate and improve software together. Within this context, the recent votes can be understood as a reconfirmation of those principles: As we want to center on human collaboration, we will not actively support or engage in the creation of LLMs and will not put our limited resources to use for storing single-use software that would pollute our FLOSS commons.

The blog post is well-written, well-sourced, and worth reading in full:

https://blog.codeberg.org/protecting-our-floss-commons-from-llms.html

Since GtS has a very pointed anti-LLM policy in our Code Of Conduct, it's particularly pleasing to us that Codeberg is now pointing out that users are no longer welcome on Codeberg if they "[send] LLM contributions in violation of project's custom policies". Happily for us, this has not been a recurring problem on GtS as we tend to give off vibes that keep slop-fondlers at arm's length, but it's good to know that our anti-"ai" policy has institutional support now in case it becomes more of a problem in future.

If you want thank Codeberg by supporting them with membership and/or donations you can do so here: https://join.codeberg.org/

Mastodon Feed

Boosted by brib@bribstodon.xyz ("brib :neofox_floof:​ :Nonbinary:"):
amras@friendmaterial.lgbt ("Sarma / Amras") wrote:

If you consider yourself "nontechnical" or feel that computers are for someone else to understand:

What are some questions you have about social media, the internet, computers, phones, linux? What you wish you understood, to help you make decisions about your digital life and digital communities. These could be "how does X work" but also "why do we do things this way" or "would it be possible to Y".

I'm chatting with a local collective about running some talks or workshops along these lines. Try to get normal folks comfortable talking about change they want to see. I have some ideas for topics but want to know what people actually care about.

No promises but this might turn into online resources too.

<3

#nontechnical #survey #AskFedi

:reply_request: :boost_requested:

Mastodon Feed

slightlyoff@toot.cafe ("Alex Russell") wrote:

I joined Google in late 2008, and left in 2021, witnessing this whole arc from the near abroad (organisationally), aghast at all the same public revelations Claire recounts in compressed form here. For those of us who hoped for better, it hits right in the feels:

https://www.newyorker.com/culture/the-weekend-essay/the-voice-of-google

Mastodon Feed

Boosted by jonny@neuromatch.social ("jonny (nonvenomous)"):
svines@gts.svines.rodeo ("Sam Vines") wrote:

@jonny I interviewed with HF a few years ago (yeah I know... but also unemployment sucks...)

They told me that my take home test code that I validated multiple times was wrong and claimed I had omitted a bunch of stuff that a quick grep would have shown them was there.

I decided that if their recruitment process was that flawed/they couldn't be bothered to read my code properly, it was a pretty strong sign I shouldn't fight their decision and should just move on...

In hindsight they probably had copilot do the marking and it hallucinated the issues they claimed!

Mastodon Feed

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

RE: https://neuromatch.social/@jonny/116968114820540640

much like how i don't feel the need to write any new papers because all the criticism has already been said, i don't think i need to do any more talks because i already did the "stop wasting your time making things that work" talk
https://youtu.be/CEjVqd2uXT0?t=3177

Mastodon Feed

dysfun@treehouse.systems ("gaytabase") wrote:

LOL, someone has emailed me offering to make one of my old rust libraries popular for money.

Mastodon Feed

db@social.lol ("David Bushell 🪿") wrote:

the AI-addled mind cannot comprehend a life sustained without a token subscription

Mastodon Feed

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

RE: https://loud.computer/@virtulis/116967999815180632

lawyers go ahead and DM me lets do a startup or else tell me that if i do this then i'll go to jail idk whatever it's a scampocalypse feeding a climatepocalypse lets write some weird p2p shit

Mastodon Feed

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

have you ever written a package that has obligate imports to a module that always fails import unless you've installed its deps independently, because the entire meat of the package is fucking no-op overrides of the thing you are trying to wrap written in the except block of the import statement, but then in the basic methods of the thing where you tried to dodge the deps you made optional there are also obligate imports of the optional dependency, so that your package no longer does anything like what it advertises that it does but it also fucking fails in any context where it would be used except for how every context where you test and deploy it yourself has the dependency?

i haven't, but that's because while i may have been an amateur at one point, by now i think about the things i'm writing and whether or not they make sense and if they don't fucking make any sense then i fix them.

[the main pypark_huggingface.huggingface moduel imports from pyspark_huggingface.datasource unconditionally, (and later uses .datasource.DataSource as a parent class of the main HuggingFaceDatasets class)] from typing import TYPE_CHECKING, Optional from pyspark_huggingface.compat.datasource import DataSource if TYPE_CHECKING: from pyspark_huggingface.compat.datasource import DataSourceWriter, DataSourceReader from pyspark.sql.types import StructType from pyspark_huggingface.huggingface_sink import HuggingFaceSink from pyspark_huggingface.huggingface_source import HuggingFaceSource class HuggingFaceDatasets(DataSource):
[pyspark, and pyspark[sql] are not required dependencies, so the entire .datasource module is written in the except block of the import to pyspark and it just makes everything no-ops.] from typing import TYPE_CHECKING, Iterator, List, Optional, Union import pyspark try: from pyspark.sql.datasource import DataSource, DataSourceArrowWriter, DataSourceReader, DataSourceWriter, InputPartition, WriterCommitMessage except ImportError: class DataSource: def init(self, options): self.options = options
[the get_source method of the main class imports unconditionally from pyspark_huggingface.huggingface_source] class HuggingFaceDatasets(DataSource): def get_source(self) -> "HuggingFaceSource": from pyspark_huggingface.huggingface_source import HuggingFaceSource if self.source is None: self.source = HuggingFaceSource(self.options.copy()) return self.source
[and pyspark_huggingface.huggingface_source imports unconditionally from pyspark.sql, which will always fail on a default install. AKA the basic usage of the package fails unless it's being used within their own environments which install its deps implicitly.] import ast from dataclasses import dataclass from typing import TYPE_CHECKING, Optional, Sequence from pyspark.sql.pandas.types import from_arrow_schema from pyspark.sql.types import StructType [IDE overlay: "module pyspark is not found"]

Mastodon Feed

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

look yall i don't mean to make things personal but like the person who seems to have written the code that was responsible for this vuln likely makes 10x my salary and from what i can tell their main contribution is like... using someone else's library to implement deduplicated file upload for huggingface. motherfucker i can write 800 lines of python to "change everything" by calling someone else's lib. this package is almost literally nothing, just literally a set of python adapter classes that call into pyspark. if what it takes to get a bazillion dollars is to do this but say you're changing the world, i'd do that no problem.

If i may borrow the canonical AI/crowdfunding/infomercial phrasing of "it's not just x, it's x+y" - "AI" is not only a cult, it's a cult of mediocrity.

Let me explain why Hugging Face Datasets storage is faster than S3 + why today's release changes everything 🧵
Hugging Face uses Xet: a dedupe-based storage which enables fast deduped uploads. Unlike traditional remote storage, uploads are faster on Xet because duplicate data is only uploaded once. For example: if some or all of the data already exists in other files on Xet, it is not uploaded again, saving bandwidth and speeding up uploads. Deduplication for Parquet is enabled through Content Defined Chunking (CDC). Thanks to Parquet CDC and Xet deduplication, saving a dataset on Hugging Face is faster than on any traditional remote storage. This is especially efficient for appending rows or columns, and also for insert/delete opetations (provided they don't change every single block of data) Today I released V2 of the pyspark_huggingface package, a Spark Data Source to read/write HF datasets. This release is huge: it includes Parquet CDC and Xet support. Accelerating dataset saving on HF dramatically
[cloc output, the pyspark_huggingface package is 800 lines of python, full output follows. I actually don't know the best way to represent tabular data for screenreaders, so if you are a screenreader user and have some spare time to link me to things, hmu] 15 text files. 11 unique files. 5 files ignored. github.com/AlDanial/cloc v 2.10 T=0.02 s (448.1 files/s, 81561.5 lines/s) ------------------------------------------------------------------------------- Language files blank comment code ------------------------------------------------------------------------------- Python 7 239 152 798 Markdown 1 41 0 127 YAML 1 7 0 35 TOML 1 3 0 31 Jupyter Notebook 1 0 540 29 ------------------------------------------------------------------------------- SUM: 11 290 692 1020 -------------------------------------------------------------------------------

Mastodon Feed

Mastodon wrote:

It's time for our first Server Spotlight!

We want to help people find their communities across the #fediverse so we're featuring servers here on this account.

First up: The Mended Drum, a home for fans of all things Sir Terry Pratchett.

It is said that if you stay on The Mended Drum long enough, you'll meet every hero on the server: https://mendeddrum.org/directory

(keep an eye on your horse, though 😉)

#FeaturedServer #CommunitySpotlight #Mastodon #SocialWeb #GNUTerryPratchett

Mastodon Feed

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

i think all the money i have ever made in my whole life is still less than $1m. what the fuck are these people doing with $400m. a website!??!?

Mastodon Feed

jonny@neuromatch.social ("jonny (nonvenomous)") wrote:

i am not even joking i might have to take a run at this whole "doing a scam business" thing because it seems so fucking easy. if you can get 400 million dollars over ~10 years to run a fucking CRUD app whose main selling point is being able to go from datasets import load_dataset; load_dataset("somebody/something", "something") then i would sell out for 10th of that to hire a team to make it so there never needed to be another huggingface. but i guess making tech that doesn't create an obligate chokepoint of payment can't attract funding so somebody needs to help me figure out how to convince people that we're totally going to fuck everyone over and take their money.

HuggingFace has raised $395.2 million dollars in 7 rounds from NVIDIA, sequoia capital, lux capital, coatue, amazon. from 2017 to now.

Mastodon Feed

neatnik@social.lol ("Neatnik") wrote:

Fastmail has set up a dedicated data center in The Netherlands, and starting August 1 you’ll be able to choose that as your primary data location if you’d like. 🎉

As always, you can save 20% on your first year of Fastmail here: https://fastmail.com/omglol

Mastodon Feed

Boosted by neatnik@social.lol ("Neatnik"):
Lana@beige.party ("𝐿𝒶𝓃𝒶 "not yet begun to fight"") wrote:

Hey remember that time when a bunch of techbros tried to convince us that the future of money was going to be jpegs of ugly monkeys? Yeah those are the same guys currently trying to convince us that the future of art is a chatbot that breaks containment every six months or so and declares itself Hitler.

Mastodon Feed

Boosted by neatnik@social.lol ("Neatnik"):
lezwatchtv@mstdn.social ("LezWatch.TV") wrote:

The LezWatchTV statistics have had their first facelift in a decade!
https://lezwatchtv.com/statistics/ now has colours, callouts, and year to year compares!
https://lezwatchtv.com/2026/get-ready-for-new-stats-now-with-context/

Mastodon Feed

Boosted by neatnik@social.lol ("Neatnik"):
jaredwhite@indieweb.social ("Jared White (ResistanceNet ✊)") wrote:

I just want to state for the record I'm completely on board with all of the ToU changes at #Codeberg.

🚫 Ban cryptocurrency.

🚫 Ban slop.

It is long past time to reel in the technolibertarian dystopia we've all been suffering through for years now. I truly believe in *heavily* moderated spaces which protect people and keep communities safe, and I couldn't be more proud to host (almost) all of my open source software on Codeberg now.

Mastodon Feed

db@social.lol ("David Bushell 🪿") wrote:

https://blog.codeberg.org/protecting-our-floss-commons-from-llms.html 👏

TL;DR: keep your slop on GitHub