HNHacker News
TopNewBestAskShowJobs

bondarchuk

4,345 karma · joined May 30, 2020

submissionscomments
bondarchuk··on History LLMs: Models trained exclusively on pre-1913 texts
>Historical texts contain racism, antisemitism, misogyny, imperialist views. The models will reproduce these views because they're in the training data. This isn't a flaw, but a crucial feature—understanding how such views were articulated and normalized is crucial to understanding how they took hold.

Yes!

>We're developing a responsible access framework that makes models available to researchers for scholarly purposes while preventing misuse.

Noooooo!

So is the model going to be publicly available, just like those dangerous pre-1913 texts, or not?

bondarchuk··on Firefox will have an option to disable all AI features
Video games are incredibly popular and my mom plays them, does that mean Firefox should have video games baked in at the base layer?
bondarchuk··on Beginning January 2026, all ACM publications will be made open access
>The problem is with any sort of closed system, the signal immediately gets gamed.

I agree, but..

>Therefore, the open system is the least bad of the available options.

this does not necessarily follow.

>A journal could still achieve prestige by curating and selecting the best available studies and research

See, this is just the kind of thing that I think will just not work when organized top-down like that. "Oh, we'll just make a prestigious journal by only letting the best papers in" - everyone could say that, but what would induce the authors of the best papers to submit them to your specific journal at all in the first place? Currently it's the fact that it's already prestigious, and this reputation has grown over many years through informal social processes that are very hard to codify.

>Scientific publishing is free speech. As such, it shouldn't be under the thumb of institutions or platforms that gatekeep for profit or status or political utility or any of a dozen different incentives that will fatally bias and corrupt the resulting publications.

Of course I agree, just to be clear I am a great proponent of openly accessible science - just think the prestige thing is an interesting corner case.

bondarchuk··on Beginning January 2026, all ACM publications will be made open access
I think the difference is with a journal like Nature people are competing for strictly limited real-estate. The famous academic could still sign however many papers they like..
bondarchuk··on Firefox will have an option to disable all AI features
For example at the moment multi-account containers is a plugin. I needed it and installed the plugin and it's fine.
bondarchuk··on Please just try HTMX
I don't think it's fine, if even the "just fucking use htmx" page can't be arsed to just actually fucking use htmx then something must've gone wrong somewhere.
bondarchuk··on Beginning January 2026, all ACM publications will be made open access
Well, yes, this is exactly the kind of well-intentioned technical solution that just will not work at all when it comes in contact with human nature. "Oh boy my paper got accepted in Nature!" vs - "oh boy some people on the internet signed my pgp thing!". Just not the same.
bondarchuk··on Beginning January 2026, all ACM publications will be made open access
I think the big missing thing in any proposed or actual fully open system is it does away with the difference between "prestigious" and "non-prestigious" journals. "Prestigiousness" is actually a really useful signal and it seems really difficult to recreate from the ground up in an open and fair system. It's almost like "prestige" can only emerge in a system of selfish/profit-motivated actors.
bondarchuk··on Please just try HTMX
>The "server" (mocked client-side for this demo

Hmmm.... I wonder why that is......

bondarchuk··on Classical statues were not painted horribly
So let's introduce a bias then, who cares? It's not a mortal offense. It would be cool to see statues painted realistically and non-horribly. And as TFA notes we have frescoes, mosaics, encaustic portraits etc.. that could be used as a guideline.
bondarchuk··on Classical statues were not painted horribly
So where are all the plausible painted statue reconstructions?
bondarchuk··on Quill OS: An open-source OS for Kobo's eReaders
>For the speed issue, if it's limited by the time to render a page

It's not, it's a physical limitation of the e-ink screen.

Any e-reader I've seen does full refresh every n pages where n is user-settable.

bondarchuk··on Germany's train service is one of Europe's worst. How did it get so bad?
You are trying to add a lot of nuance to a topic which is just not very nuanced. Trains in germany are a nightmare because of decades of mismanagement. All the rest is copium, as the kids say.
bondarchuk··on Germany's train service is one of Europe's worst. How did it get so bad?
It's really a pity for a large part of international train travel too, given Germany's central position in Europe. Many people I know (moi incluis) would really like to take the train for holiday or work travel (instead of airplane) but it's just not worth the risk of having to deal with Deutsche Bahn. Holland to Denmark by train, tried it once, never again.
bondarchuk··on Berlin Approves New Expansion of Police Surveillance Powers
>Mothballed and Milton are describing how they think things ought to be, or inevitably will be

There is again an enormous difference between describing how things ought to be and how things inevitably will be.

>for you to argue that it's somehow not real

Noone was saying that.

bondarchuk··on Using secondary school maths to demystify AI
I wish I would've learned about ANNs in elementary school. It looks like a worthwhile and cool lesson package, if only they'd do away with the idiotic dogma...
bondarchuk··on Using secondary school maths to demystify AI
>Consider for example Orch OR theory

Yes, or what about leprechauns?

bondarchuk··on Berlin Approves New Expansion of Police Surveillance Powers
"about to learn the hard way to put them back" would seem to imply you expect some real consequences from the lack of "fences".
bondarchuk··on Berlin Approves New Expansion of Police Surveillance Powers
Is-ought distinction. Mothballed and Milton are describing how things are, not how they should be, which latter seems to be your interpretation.
bondarchuk··on Berlin Approves New Expansion of Police Surveillance Powers
That explanation would only hold water if we were dealing right now with all kinds of problems stemming from lack of authoritarianism and too much freedom. I'm not really seeing it but maybe you have something in mind.
bondarchuk··on Berlin Approves New Expansion of Police Surveillance Powers
Just some data to give a little bit of context, however flawed or reductive it may be..

https://en.wikipedia.org/wiki/List_of_freedom_indices#List_o...

bondarchuk··on UK House of Lords attempting to ban use of VPNs by anyone under 16
You can just read it on Wikipedia https://en.wikipedia.org/wiki/Censorship_in_the_United_Kingd...

("(Redirected from Freedom of speech in the United Kingdom)" kind of funny innit)

bondarchuk··on Cassette tapes are making a comeback?
It's just annoying to post unverifiable numbers without a credible source and expect others to do the hard work of verifying it (or just take you at face value, probably). It has nothing to do with AI except insofar as your own feelings on AI convinced you this is a reasonable method of communication.
bondarchuk··on Slop Detective – Fight the Slop Syndicate
Radical solipsism, the ultimate philosophical cop-out...
bondarchuk··on Cassette tapes are making a comeback?
>Edit: according to gemini AI:

>* Type I had a dynamic range of about 50bB (roughly 8 bits)

>* High quality tape with Dolby B, C and dbx yielded roughly 65, 75, and 85dB SNR (about 11, 12.5, and 14 bits)

>So you could get pretty close to CD quality, but not quite.

Source? AI content without it is less than worthless.

bondarchuk··on Slop Detective – Fight the Slop Syndicate
But you have to engage with it before you can find out whether you like it.

>If you can't tell the difference, and it's entertainment stop worrying about it.

At the end of the day it's a philosophical/existential choice. Not everyone would step into the awesome-life-simulator where you can't tell the difference. On similar grounds one might decide on principle to consume only human-made media, be a part of the dynamical system that is real human culture.

bondarchuk··on CoMaps emerges as an Organic Maps fork
Instead of genetically engineering humans to change their behaviour, I think forking the project is a good solution.
bondarchuk··on The surprising benefits of giving up
A nice article on the benefits of giving up from a magazine promoting religion under the guise of science.

https://www.templeton.org/grant/nautilus-magazine

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

bondarchuk··on Windows 11 adds AI agent that runs in background with access to personal folders
That was WordPad, right? Except it also loaded instantly, probably.
bondarchuk··on Transpiler, a Meaningless Word (2023)
Whenever someone argues the uselessness or redundancy of a particular word we just have to remember that the word exists because at least two parties have found it useful to communicate something between them.
← PreviousPage 8 of 34Next →