The image displays a dark blue background with a search engine interface. At the top center, the word "kagi" is prominently displayed in white lowercase letters. To the right of the text, there is a cartoon illustration of a dog with long ears, labeled "Kagi Dogge by Lynn." Below the logo, there is a search bar with the placeholder text "Let's fetch..." in orange. To the right of the search bar, there is a "Search Options" link. At the bottom center, there is a blue "Search" button. The bottom right corner contains the copyright notice "© Kagi. Humanize the Web."
―
The image displays a website interface with a minimalist design. At the top center, the logo "kagi" is prominently featured, accompanied by a cartoon dog character named "Kogi Doggo by Ludo," sitting next to a yellow ball. Below the logo, there is a search bar with the placeholder text "Let's fetch..." and a search button labeled "Search." The search bar includes a "Search Options" link on the right. The background is a light gray color, and the overall design is clean and modern. At the bottom of the page, there is a navigation menu with links to "About," "Blog," "Changelog," "Live Stats," "Swag," "Privacy & Terms," "Help," "Feedback," "Support & Community," and a copyright notice stating "© Kagi, Humanize the Web."

Provided by @altbot, generated privately and locally using Ovis2-8B

🌱 Energy used: 0.606 Wh

I see only two possible things a court could decide, neither are good for AI companies:

• AI-generated works are not creative works and are therefore not subject to copyright (seems to be the prevailing opinion).
• AI-generated works are derived works of the training set and are therefore subject to all license restrictions on everything that went into the training set.

The second would be much worse for AI companies (all outputs would be subject to copyright, but not owned by the end user or the model provider unless they have explicit license terms for everything used to train the model).

I would expect the second to be more consistent with copyright law. A poor-quality JPEG of a photo of a painting is a derived works of the painting and not independently copyrightable. Generative AI is lossily compressing a load of things and then running the decompresser with a seed value. I doubt anyone taking these things to court so far wants to argue that side though.

You can track the SocialCG's public discussions at w3.org/wiki/SocialCG - it has links to the email list and meeting notes. Here's an example of somebody from a Meta-funded non-profit (who's got a major role in the discussions going forward) attempting to shut down criticisms of Meta on the email list. lists.w3.org/Archives/Public/p…

The SocialCG is currently in the process of coming up with a charter for a Working Group to create the next version of the ActivityPub spec. Things to look for: does Meta have a representative on the working group? (Since they're a dues-paying W3C member, they're entitled to one, so presumably the answer will be yes.). How many people from Meta-funded non-profits (or organizations that take funding from Meta or Meta-funded non-profits) will be in the group? What roles will they play, publicly and behind the scenes? How many people critical of Meta will be involved -- and what roles will they play?

@nemesis @adrianmorales @ophiocephalic @ntnsndr

@Mik3y indeed, openSUSE is probably the best option if this is important to you (and I did note that in the article actually), but this is probably something I’d be less concerned about for all the reasons @niels shared.

We have a pretty good sense of what’d happen if RedHat mismanaged Fedora, because a few years ago when people took issue with RedHat’s CentOS changes, the community stepped in with Rocky & Alma Linux, and people formerly on CentOS had a pretty easy pathway out. Very likely something similar would happen with Fedora if things really went badly for whatever reason.