On the closure of the "On LLM policy of HaikuPorts" thread

[kaban4ik]

I apologize if I wrote too emotionally and offended anyone, the main message is: People, we need to talk and create compromise solutions that are convenient for everyone.

[us3r1d]

@waddlesplash

that means we’re actually moderating less than we would under normal circumstances

This is certainly true.

There are several posts in the last day in this thread that I would have at the very least contacted the author and requested that they remove the personal attacks from; there are a few I would probably have hidden until they were rewritten.

But as you say, when things get all dramatic and tempers are already high that can just make things worse.

[cian]

The “owners” of BeBits and Haikuware had become one single person by that stage. Both sites were going to go together if he took offence to something, and based on dealing with him before packagefs, he was going to take offence to something eventually.

I also find packagefs to be overly complicated and difficult to work with, but my minor enough contributions to beports/haikuports had long since been overtaken by other, more talented people so there wouldn’t have been any benefit for me having a hissy fit and going away anyway!

[Sebrof]

Polls are like referendums. They are fine for deciding matters that aren’t important - like choosing a national anthem.

They are disastrous when used to decide important and complicated questions, of which the most glaring recent example was the catastrophic decision, decided by a referendum, for the UK to leave the EU.

I know very little about LLMs, and there may be people who know even less, but in a poll we are accorded the same importance as people who know a lot. And it is the latter who should be making the decision, not an unregulated consituency of people whose competence to vote sensibly on this is entirely unknown.

So I think the mods are right to ignore this poll. And indeed all polls.

[Akakor]

Don’t start a poll if you’re not willing to accept the outcome. The majority of the voters are fine with LLM use, but since a few moderators and devs (a minority of the voters) are against, the outcome is clear. Apparently, not every vote weighs the same.

[us3r1d]

@Akakor

The majority of the voters are fine with LLM use

Actually, as it stands right now 58% of respondents to that survey stated that they wanted to either ban or discourage LLM use for translation on this forum. 42% said that it should be allowed with no commentary on it.

84% did say it should be allowed for translation on this forum, but half of those also said that its use should be discouraged in favor of non-LLM translators.

So, adding a statement to the forum policies that LLM-based language translation is allowed but we would prefer that users employ other translation technologies would be exactly in line with the current results of that survey.

[SURVEY] Policy proposals for LLM translators

All right, it seems discussion about LLMs for translation has mostly died down. Let’s see how the community at large feels about each of the proposed policies. First, which policy would you prefer? poll Second, which policy do you like the least? poll Now, how much do you like/dislike policy 1, which bans LLMs for translation? poll How much do you like/dislike policy 2, which allows LLMs for translation? poll How much do you like/dislike policy 3, which allows LLMs for translation, but enco…

Also, it’s important to remember that survey is only about LLM-based translation on the forum; it doesn’t ask about any other LLM use or LLM use in general.

[Sebrof]

@Akakor

Apparently, not every vote weighs the same.

And that is right.

[donn]

@Akakor

The majority of the voters are fine with LLM use

… for translation of a forum post into English.

Sebrof’s right, policies that are more critical, and technical, should certainly not be decided by a poll of random voters.

What I hate about threads like this, is seeing people here who play an important role in Haiku development, wasting their time and enthusiasm on this endless mudpit. We have picked up on the fact that there’s some trolling going on, but really, trolls aren’t specially inimical individuals, they’re all of us, when we lose sight of what’s productive exchange.

[suhr]

84% did say it should be allowed for translation on this forum, but half of those also said that its use should be discouraged in favor of non-LLM translators.

I should point out that encouraging using non-LLM translators is not the same as discouraging using LLM translators. The pool variant clearly says “encourage” rather than “discourage”.

[PulkoMandy]

@X512

And make compromises instead of pushing packagefs at any cost, for example installing HPKG to packaged directory instead of introducing non-packaged directory, so classic ZIP extract install will still work. Keeping common directory.

Oh god, we’re having that debate about the package manager again? It’s been more than 10 years. I am still waiting for a single example of an application that was broken by this. My offer to investigate and fix these issues still stands. Over these 10 years, no one has been able to provide one single app to fix. I cannot fix imaginary problems that don’t actually exist.

[Pap]

@nipos

These flamewars which start again and again and again only show that it seems impossible to make compromises with LLM fanboys because nothing seems good enough,and when they cross the line and moderators have to take action,the next flamewar is guaranteed to start.
My suggestion: It’s really time for a hard ban on anything LLM-related.

I quoted just one important part but really, I should quote everything.

It is crystal-clear nothing will ever satisfy the LLM-lovers. Moreover, think of how many hours are wasted just in this thread — which is not the first nor the last of this kind. All such threads were started with one clear intention: to start yet another flame war, ignited by the exact same ballerinas. They will never stop, they 'll keep pushing for the “Holy LLM”, again and again for an eternity.

So yes, I think it’s time for the moderators to unearth the goddamn ban hammer and finally USE IT. LLM-lovers will call it “anti-LLM attitude”, “witch hunt”, or something similar. My response is simple: who cares what you think, either shut the hell up or get the hell out of here (I really tried hard to censor myself here). We had enough of your nonsense, go promote your LLMs prayers elsewhere.

Sorry for the language, but I really-really had enough. It’s not hatred or anything (in fact, LLM-lovers are indifferent to me). I just want to see peace again in this forum, with productive conversations and threads I could learn something. Instead, I get endless bullshit made by the same LLM zealots non-stop. It is more than obvious they will never stop and we will never find peace again in this forum as long they are around. So, please, please, please moderators, use the ban hammer. I see no other solution. You either use it, or let a handful of foolish people destroy this forum.

Now I return back to my silence.

[jackburton]

@Pap

It is crystal-clear nothing will ever satisfy the LLM-lovers. Moreover, think of how many hours are wasted just in this thread — which is not the first nor the last of this kind. All such threads were started with one clear intention: to start yet another flame war, ignited by the exact same ballerinas. They will never stop, they 'll keep pushing for the “Holy LLM”, again and again for an eternity.

Interesting that you use a strawman to show the opposing position.
There are a lot of people who aren’t “LLM-Lovers” who keep pushing for the “Holy LLM”, but just don’t oppose the usage of LLMs or any other tool.
People who think that people are free to use their preferred tools.
Almost every banter here has started when admins hid posts or when people heavily attacked other people because of LLMs usage, not by people who used LLMs.

[zuMi]

@Pap

who cares what you think, either shut the hell up or get the hell out of here

Not a Voltaire fan, I see

[x68k]

Did Voltaire argue that you can’t tell a barking dog to be quiet, because the dog has a right to free speech? If he had lived today, would he have argued for robot rights? I think it’s safe to assume that he was and would be talking about humans only.

[nephele]

@jackburton

Almost every banter here has started when admins hid posts or when people heavily attacked other people because of LLMs usage, not by people who used LLMs.

Uhh, no. The first topic started when OP attacked the community for “silencing AI tools”, for no reason… this particular topic was started for “how dare waddlesplash add moderation context” etc. Considering this topic had no actuall critizism on the moderation it is hard to see these as anything than bad faith attempts to derail discussion and create division. It’s perfectly possible to discuss these technologies in a reasonable manner without dishonest and hostile framing.

[zuMi]

@x68k

Did Voltaire argue that you can’t tell a barking dog to be quiet, because the dog has a right to free speech?

Are you comparing users to dogs?

[zuMi]

@nephele

Considering this topic had no actuall critizism on the moderation

my previous messages are all requests for clarification on moderation (as indicated by the thread topic: anyone who started talking here about LLM went off-topic). Therefore, I kindly ask you to take another look and try to give me an answer on recent events.

[PulkoMandy]

@waddlesplash

Hmm, where was this? Perhaps my memory is wrong… I just searched the mailing list archives and my own email inbox and couldn’t find where this happened, though.

The closest I can find is I decided to go away from forum - #35 by waddlesplash

@X512

Discussion for my inclusion to Haiku developers with commit access started on mailing lists without any my request and then PulkoMandy said, that he reject my candidate for not following coding style.

I cannot find anything on the mailing list. As you say in the next sentence, there was at some point some discussion “behind the scenes”. We do that to avoid the embarassment of starting a public discussion (mailing list or even worse, the forum with hundred of users) and then having people voice concerns publicly, which is not great either. Is there a correct solution here? So that’s why these things are discussed in private first.

Also, I did not reject anything on my own, I voiced my opinion and concerns about your contributions at the time (probably about the RISC-V port). These concerns would not block your application, and if other people don’t agree with my too strict opinion, either they would convince me, or there would be a vote on the mailing list and I’d lose it.

Also, you reduce it to “coding style”, but my concern was more that you sent a lot of code for the RISC-V port that required a lot of subsequent refactoring and cleanup, allowing the changes to benefit other parts of Haiku. Indeed I think some of these changes are not quite done yet (and are not easy to do), such as the PCI initialization and resource allocation which we recently discussed. I also understand that we have different approaches to things and yours is not worse than mine, just different. I hope I can be open about that and we are not in a place where being critical of other people is disallowed. Because that’s how I learn things: by people pointing out my mistakes.

Things have moved on since then, and my “maybe not yet” concern from 5+ years ago is now obsolete, as a lot of the communication difficulties have been resolved, both on a human level (we know and understand each other better ow than then, I think) and on the technical level (at the time you didn’t really want to get into mailing lists which were still used to run the project, now it looks like pretty much everything moved to the forum and the mailing list is mostly unused).

You would not be the first person to whom this happened. Indeed, waddlesplash got rejected several times before he was granted commit access.

[3dEyes]

@Pap

So yes, I think it’s time for the moderators to unearth the goddamn ban hammer and finally USE IT. LLM-lovers will call it “anti-LLM attitude”, “witch hunt”, or something similar. My response is simple: who cares what you think, either shut the hell up or get the hell out of here (I really tried hard to censor myself here). We had enough of your nonsense, go promote your LLMs prayers elsewhere.

I would like to remind you that among those who disagree with your stance are quite a few people who have made immense contributions to Haiku and have been with this project for two decades. Are you suggesting they should leave too? Should they just walk away? Or should they leave and take with them all the future contributions they could still make? Don’t you think that in your “holy war”, you have forgotten the main reason why we are all here?

[X512]

@PulkoMandy

Also, I did not reject anything on my own, I voiced my opinion and concerns about your contributions at the time (probably about the RISC-V port).

If I remember correctly, is was much earlier than my RISC-V port.

I have impression that Haiku code review process is too focused to code style check rather then actual functionality and correctness checks. I have seen multiple commits that were looking nice, but caused serious regressions that stayed for half-year and more, mostly Tracker changes by John Scipione. Even now I see recent libtracker regressions that caused file open/save dialog to freeze/crash application (#20125).

I think that more attention must be applied to checking code correctness during review rather than counting characters per line or class tab formatting.

Tracker/Deskbar is based on original open-sourced BeOS code, so in theory it should be already Haiku R1 ready. I think that code review for Tracker/Deskbar changes should be especially strict and refactors to make code “look nice” or questionable new features should be generally rejected.