This is an especially disingenuous argument. Being anti-Big Tech in general is a separate issue and would have to be discussed with the people who actually develop Haiku, in terms of cost-benefit (where the costs are social). In the specific case of AI, there’s no way to use one ethically, no matter the source. The costs are unacceptable regardless.
Besides, we don’t have to personally fix all the world’s evils. But there are things we can do in the moment, such as not causing even more harm.
HaikuPorts started years before Microsoft bought github in late 2018.
But, indeed, moving away from github is in itself a topic, but a different one.
The costs are unacceptable to you.
If you really mean all of what you say, please close your laptop, stop whatever you’re doing and stop living, because almost everything you have or do in your western, modern life, has a negative impact on the planet and someone else, living in another country, producing that article for you.
May i remind everyone that this is a meta topic on moderation principals and not a discussion on LLMs and the project’s stance on them.
As a moderator of a large community, topics are often locked when discord gets out of hand, and often a message pertaining for the reason of the lock is posted. THIS IS COMMON PRACTICE. Nearly every public digital forum works the same way.
Regarding the topic this thread has degenerated into…
LLMs are a tool. Tools can be useful. Whilst i would not like to see large quantities of code in Haiku’s main tree generated by AI, i think AI CAN be used to assist development. However, regarding 3rd party software, i don’t think Haiku really has jurisdiction on regulating what can and can’t be allowed purely on how much of it was AI generated. Will it be included in the base system image? No. but there should be an option to download and install it if you wish. Restricting software on political ideologies will lead to significant fragmentation, and the great thing i find about Haiku, is it’s never (yet) suffered such fragmentation.
“Yet you live in society. How curious. I am very intelligent.”
Congrats on proving me extra right on all accounts. This is exactly the attitude I’m talking about.
The impact of the LLM issue on the community, and how that should be handled by moderators, is exactly the topic at hand. You can’t divorce the technical arguments from the social context. And that’s all of us.
Of course, that’s just my personal opinion… you and the others, stop babbling about pollution, antisocial behavior, and other nonsense. I don’t think you share the same opinion about capitalism, which creates everything you talk about anyway. Your position seems childish to me, to say the least!
Tbh these are valid points to be concerned about. Genuine question though to you and everyone else who shares the same concerns, would you be more open to adopting LLMs if the contributors were using locally running open-source LLMs trained on the publicly available datasets?
Genuine question though to you and everyone else who shares the same concerns, would you be more open to adopting LLMs if the contributors were using locally running open-source LLMs trained on the publicly available datasets?
The last requirement limits the usefulness of available models quite a bit, and I’m not sure if anyone does this in practice.
Why do you think so? Have a look at StarCoder2 models and related Stack 2 datasets: bigcode (BigCode).
This illustrates the point, 15B is the largest model size of StarCoder2, while you can run bigger models on a good hardware locally. It also seems to be very old, so it’s not trained for reasoning or tool usage.
I think you can remove the problems about massive web scrapping and very high resource usage (or at least have poeple pay this price locally as they run the LLM, making them heat their own house with their own computer). It may also be actually open source (not just open weight or other not-quite–free-software schemes).
It doesn’t remove the other aspect of this: if people solve problems by talking to their LLM instead of talking with other humans, I don’t really like the type of society that this is building, which will be more individualist, and hide from people the fact that they rely on a lot of other people to have things running. It then creates a fertile ground for libertarianism, built on the fake perception that you don’t depend on anyone else.
I think this doesn’t really change if the LLM is locally trained. But also, that the LLM here is just a consequence of a pretty broken organization, and not really the cause. We live in a world where people prefer to talk to a machine that is wrong some significant percentage of the time, rather than talk to other humans. And, well, if it comes to that, I hope I can say that Haiku at least tried to push back against that.
I would rather have a discussion about how we can make this space more human friendly, and not about how to avoid human interactions by using LLMs because some people have come to the conclusion that this is more efficient/productive (hint: you will not be remembered for the number of lines of code you wrote, but you will be remembered for your interactions with other people) or more safe/less toxic (hint: we should kick out the humans or the behaviors which make the place toxic, instead, if there are some).
That would be the obnoxious, intellectually dishonest people who demand that a space be carved out for them, and answer all the common-sense, well-supported arguments against it by stomping their feet and crying, “but I want to use my shiny toyyy!”
Hint: this is the politics that our moderators asked us not to discuss for years, coming back to bite us in the behind, predictably.
Please do the right thing now. Please.
Exactly, no politics here… this was the rule and hypocrisy is just around the corner.
Ironically, you just threw a political view in your previous comment yourself.
So, welcome to the hypocrisy club, it’s a very large club, and I’m also a member like anybody else.
The only thing you are proving by your comment is your own hypocrisy: AI ruins the planet and should be banned from Haiku, but everything else you do or have, does not have a negative impact on people and planet. You’re a saint and AI users are evil.
When you exagerate anyone’s opinion, it becomes ridiculous. Who would have thought?
The point is: we did without AI and LLMs for several decades at this point. We can do without them. And the benefits (apparently, writing code faster, but not necessarily better) may not be worth the extra cost.
The balance in other cases is different. I think my bicycle from 1998 has paid off its carbon offsets. I had to buy a “new” phone (from 2021) recently because the old one fell and broke (otherwise it would have lasted me a few more years). In these case, I think the benefits are worth the cost, otherwise I wouldn’t have bought such things.
That’s actually a good question and yes,I think this would solve a number of major issues with LLMs.
As PulkoMandy already pointed out,there would still be the social issues of replacing communities and having developers stop talking to each other,or users stop talking to developers.
And it clearly wouldn’t solve the quality issues,the fact that LLM-generated code often has bad quality and just doesn’t work,or even worse,has subtle issues that you only discover much later.
But it would solve the major ethical and legal issues that make that technology fundamentally flawed and makes me not want to have anything to do with its output.
Big Tech companies generally share a mindset that the law doesn’t apply for them.
They do obviously illegal things if they think the fine will be cheaper than the profit/marketshare/whatever they gain by doing it anyway,or they think there will be no fine at all,because they’re too big to fail.
Even if most copyright cases aren’t decided yet,the fact that there are so many of them clearly shows that LLM companies think copyright doesn’t matter.
That’s additionally to the fact of DDoSing free software projects infrastructure or even random websites that I brought up multiple times.
So yes,if a variant of that technology is used that is not built on breaking copyright or damaging other peoples infrastructure,it would be a lot easier for me to tolerate,even if there’s no chance I’d use it myself.
Much like AI shills share a mindset that they’re entitled to do whatever they want, and the rest of us are big meanies for not letting them. Go figure.
Hi, folks.
Y’all very clearly want to get off into the weeds arguing about the merits or lack thereof of LLM.
Fine. 
But that is off-topic for the Haiku forums, so I’m moving this part of the thread to Off-Topic.
It’s getting a little heated, though; I’ve looked through past moderation actions here, and I don’t think anyone in this thread has definitively crossed the line into needing to be punted to cool off yet by the standards this forum has used but a few of you have come very close to it.
So, please,disagree but do it in good faith and don’t make assumptions about the motives of the people you’re disagreeing with.