We’d like to remind everyone of the importance of following our community guidelines to ensure our forum remains a welcoming and productive space for all participants.
Key points to keep in mind:
Give space for others: We encourage everyone to contribute, but it’s equally important to allow others the opportunity to share their perspectives. Please be mindful of the frequency of your posts so that everyone has a chance to participate.
Avoid repeating arguments: To keep discussions fresh and engaging, please refrain from repeating the same points or arguments. Let’s focus on moving the conversation forward by bringing new ideas or perspectives.
Limit creating new topics: Before starting a new topic, please ensure that it’s necessary and that the subject hasn’t already been covered in an existing thread. Consolidating discussions helps everyone stay organized and focused.
Moderation Notice
Please take a moment to review our community guidelines and keep them in mind while participating. Repetitive posting, unnecessary topic creation, or failure to follow guidelines may result in actions being taken by the moderation team.
I agree that this forum is now hard to follow, and the arguments over data are now across a large number of threads. I have myself been in a position where addressing concerns becomes unwieldy. It is hard.
I have also heard OSI lament that discussions are moving elsewhere, to LinkedIn and so on, which is rightly a concern of OSI. I don’t mind OSI trying to direct the conversation into a particular forum, and recognize that the OSI must ensure that the forum is functioning well.
However, I have noticed that OSI people aren’t engaging that much in this forum anymore, for the most part, OSI seems to respond in blog posts. The level of engagement from OSI is quite visible in the usage statistics. That too, would be a force that drive people away from the forums, probably at least as much as the proliferation of topics and the move to LinkedIn. People get frustrated and forums start deteriorating when they are not where discussions occur.
I certainly share most of the concerns that @Shamar listed in List of unaddressed issues of OSAID RC1, and would love to see OSI address them one-by-one. That, I believe would be very helpful and move a substantive discussion forward. I would also like to see OSI respond to @quaid 's proposal. Without such engagement, I fear appeals to community guidelines are moot, and more importantly, much of the community will be alienated if the RC is approved on Oct 28th.
I have found the forums completely unbearable the past few weeks: the very same arguments, no matter how good they are, are reposted tens of times, by the very same 2-3 people.
I understand your requests, @kjetilk, and find most of them to be reasonable in terms of accountability. But I don’t think that the lack of answers on those points justifies the previously described behavior. In fact, I consider that nothing could possibly justify it.
(And note that I even agree myself with a number of the critiques raised against OSAID RC1, but I never dared flooding the forum with my arguments/counter-arguments. Posting them once should be enough, in any civil debate.)
My 0.02€, as a mere volunteer participant in the OSAID process.
I’ve found the forums interesting and lively over the past few weeks, with quite a few reveals and unexpected developments playing out.
I see the community putting forward constructive and substantive proposals, carrying out critical analyses, sharing relevant material from beyond the forums, and engaging in respectful dialogue.
I share @kjetilk’s diagnosis that some degree of repetitiveness in the most recent week or two may be the result of issues not clearly being responded to, except perhaps obliquely in posts on the OSI blog.
With regard to the very helpful open issues thread, for instance, I found the absence of OSI engagement or acknowledgement remarkable.
The main (again oblique) response to this I see is in @nick’s admonition, above, to “focus on moving the conversation forward by bringing new ideas or perspectives.”
But that is not moving the conversation forward — it is changing the topic.
Perhaps the persistence of the can of worms that is data sharing is not to the liking of those who want to rush to a release, but it won’t be gone just by looking away.
Also having limited time, I found it both unbearable and interesting
I don’t want to provoke further metadiscussions, but I find that things go this way whenever you have an open discussion where people feel unheard. It could be a law of the information space (and we’ve not even near Godwin’s probability yet).
I think the way to go once that happens is not to try to limit the conversation, but to restructure the manner of the conversation. For example, get the open problems into issues, progress them across a Kanban board, direct the conversation so that everyone discuss the same problems at the same time in a predictable manner so that also those who do not have much time know how to prioritize their time. Discourse, however shiny it is, just cannot support conversations of such complexity as we have here.
But what if it isn’t?
What do you do if your interlocutor simply ignores the problems you point out, even though you know they will cause harm to many people?
Agreed, and while that’s often true for academia, this is a one-sided “debate” with showstopper issues being bureaucratically recorded only to be ignored, I’m told due to lack of authority. Different people naturally react differently to being ignored: some trust the process and sit on the sidelines after saying their piece, others feel it important to hold the line on principle, while many have simply given up and wandered off.
It occurs to me the OSI has never actually done anything like this before as the original OSD was a rebadged verbatim copy of the DFSG published the year prior to its incorporation. Rather than codify requirements to protect the four essential freedoms for AI by consensus, they’ve contracted it out to an experimental “co-design” process that embarked on the impossible mission of defining something (anything) “open” that would satisfy the most diverse audience possible — which often does produce better results but was not at all what was required here and now constitutes a sunk cost.
As at today, OSI Finalizes a ‘Humble’ First Definition of Open Source AI which totally redefines what “Open Source” means for everyone just as AI eats software, abdicates responsibility for risk to governments despite myriad showstopper issues, and shifts focus to “only bug fixes” despite not actually fixing any bugs (beyond dotting i’s and crossing t’s in the closed HackMD repo — bikeshed painting if I’ve ever seen it).
After the high-fives at ATO are done it’ll be like the dog who caught the car, with a less-than-useless definition and a backlog of changes that could have been incorporated at any point this year continuing to be ignored. It won’t be hyperactivity that kills this forum, but ineffectiveness. Maybe we have outgrown it and do need to move to a more traditional way of working and tracking outstanding issues, like an open Github repo that can be cloned and forked without gatekeepers. Perhaps the same can be (and has been) said of the OSI itself, and it should get back in its lane and focus on software.
When the OSI first began using the OSD, which was essentially a copy of the DFSG, it was an oligarchic organization, unlike today. OSI was originally a campaign organization for the then-unknown term “Open Source,” and I remember there were a few internal conflicts as well. However, the OSI of today is far more democratic compared to back then, and I believe it has been making significant efforts to gather input from a wide range of communities. It’s true that OSI has never undertaken something like the OSAID process before, but in the past year since I’ve been involved with OSAID, I believe the work done by OSI’s staff is more than worthy of praise.
Let me use Debian as an example. Debian is one of the great examples of a large-scale, free and democratic organization successfully developing software, but it was also a place where terrible flame wars occurred, especially about 20 years ago. The significant delays in the Sarge release symbolize that era. At that time, many people who grew tired of the endless repetition of the same arguments in flame wars fled to the Ubuntu project. Ubuntu had a Code of Conduct, whereas in Debian, even discussions to create a CoC couldn’t make much progress.
I feel that now is the time to move the discussion forward. We must avoid repeating what happened with the Debian project 20 years ago. Based on the collective input I’ve received from the AI/ML community that pursues freedom, OSPO representatives, legal experts, and the Open Source community, it seems that a majority believes the current draft’s definition for access to training data is sufficient, at least in Japan. Furthermore, embedding small AI models into various software has already become commonplace, and I’ve heard requests for a swift definition of Open Source AI due to the legal challenges around license compliance. Personally, I’m not fixated on the timing of ATO, but I increasingly feel that an early release of version 1.0 is desired.
The core of the Open Source AI Definition is the four freedoms. Everything else is secondary. If it becomes clear after the 1.0 release that these secondary aspects conflict with the four freedoms, we can simply prune those parts and, if necessary, graft on new elements. We will continue to make the necessary efforts to defend the four freedoms. As I’ve mentioned some times before, the current version of the OSD is 1.9. The release of 1.0 simply marks the beginning of a long journey.
unfortunately, the current definition undermines at its foundation a European regulation designed to protect our society and democracy from the risks posed by artificial intelligence.
and I’m not talking about fairy tales about existential risks, but concrete risks such as discrimination of minorities, unfair competition, misinformation, lack of accountability in case of damages etc.
my Portugal is a small country compared to big and rich Japan.
but the European Union has about four times the population of Japan.
the OSI itself has been transparent with respect to the role of the AI Act in justifying and initiating the process that led to this definition. but as far as I have been able to observe as a lurker in recent months, the problems being discussed have been known from the beginning and have not been addressed or solved by the OSI.
so, with all due respect, the urgency for Japanese companies to have a definition as long as it is, in the eyes of any European citizen comes second.
I am astonished and embittered by the behavior of the OSI.
and there are many who are here in Europe.
