![]() ![]() Similarly, proprietary models seem to pose significant antitrust problems, and a deeper understanding of open's management of power concentration could spur good policy proposals in that area. If open models prove that training disclosure is feasible, policy activists will be able to demand that of all models. For example, many proprietary AI vendors are taking the position that disclosing training materials is practically impossible. Policy think tanks could take this as an opportunity to study how open techniques could improve ML regulation for all models. As part of this, open ML communities need to build, and circulate, positive stories about open ML's impact, to counter negative stories about openly racist and sexist finetuning of open models. Large open players must build on the current work on security regulation (eg) to develop relevant policy advocacy skills in the open community. Or to put it another way: we cannot kill the patient (the valuable humanity-wide commons) to cure the disease (genuine problems with privacy, consent, and autonomy) what's needed is an effort to articulate this position in a way that will resonate with policy-makers. This advocacy will need to protect both the forest (the commons that society fruitfully draws on) and the trees (the creators who build the commons), a tricky but critical position. ![]() Since many of the best training sources are publicly-generated commons, there may be unique opportunities to advocate for fair usage of the digital commons and organize the digital labor that creates them. This could greatly accelerate the availability of public-benefit data sets by allowing those communities to focus on their unique data (including regulatory compliance), rather than reinventing governance wheels. Think a combination of: C4 or LAION’s data sets Wikimedia’s community building and trust and safety and Linux Foundation’s scalability and limited liability structure. There is a significant need for a public-benefit non-profit that provides standardized governance and trust and safety infrastructure to public-benefit data sets and data communities. Please let me know if you have any questions-I'm happy to answer what I can or direct you to the excellent CC team.Īnnouncing CC’s Open Infrastructure Circle - Creative Commonsīrilliant insights on open machine learning and artificial intelligence from Luis Villa: If your company benefits from these licenses (and I know many of you do!), I urge you to figure out how your company's charitable funding streams can find some pocket change for this important, long-term work. So today we've launched the CC Open Infrastructure Circle to help provide long-term, stable funding for this critical work. ![]() Unfortunately, like much infrastructure, that work is both expensive (lawyers!) and not very glamorous. What I-and CC's funders-tend to be excited about is the interesting, trendy stuff CC does, like AI and climate.īut that overlooks a really important role CC still plays: the critical work of stewarding the Creative Commons licenses, which are core infrastructure for so many important resources across the web. As some of you know, I joined the board of Creative Commons earlier this year. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |