Andreas Ahlenstorf

Today, I learnt what "sealioning" is: https://www.merriam-webster.com/wordplay/sealioning-internet-trolling Absolutely love that word. (This is not a sub-toot. I repeat, this is not a sub-toot!)

@christin@lsbt.me Inzwischen gibt es https://codeberg.org/forgejo/security-announcements/issues/56 Wäre das Problem richtig schlimm, wäre es Teil von einem Security Release gewesen und es hätte vorab keine öffentliche Änderung gegeben.

@christin@lsbt.me Ich würde ja eher die Beschreibung vom Forgejo Security Team lesen: https://codeberg.org/forgejo/forgejo/pulls/12782 Da lernt man auch, warum die Änderung nicht Teil der letzten Security Releases war.

@leanderlindahl@social.folkdata.se @Tutanota@mastodon.social Codeberg runs Forgejo (https://forgejo.org/) which can be self-hosted, too. And there are many other instances that run Forgejo, not only Codeberg: https://codeberg.org/forgejo-contrib/delightful-forgejo#public-instances

@AliveDevil@tauri.earth @forgejo@floss.social That is not possible because Forgejo supports package formats that are not compatible with that model, for example, Debian. The best option (so far) seems to be "repositories, but for packages". See https://codeberg.org/forgejo/forgejo/issues/3577

@javahippie@mainz.social Ich finde schon heute Code Review schwer erträglich und versuche, Prozesse zu nutzen, die ohne auskommen. Wenn Software Engineer zum Agent Operator wird, der nur noch überwacht, muss ich zurück zu Service Design, UI, etc. Allenfalls ginge auch System Engineer. Meine Netzwerk-Skills sind leider verstaubt, weil ungenutzt.

@Gjoel @patriksvensson Most of the things are machine learning (ML), not AI. It is absolutely crucial not to call ML AI, because the term AI causes people to assign it magic properties which it does not have. Nobody who taught me statistics, ML, or the very basics of neural networks (we didn't have DNNs yet back then) called this stuff "AI". But yes, my sample size isn't that large (three universities, lots of books).

@Gjoel @patriksvensson Spam filters of the 2000s were Bayesian statistics. Translations and TTS of the 2000s and early 2010s were Markov models. If you consider those things as intelligence, the term becomes absolutely meaningless.

@dlatchx @jsbarretto No. Projects leave issues or PRs behind for various reasons.

@Gjoel @patriksvensson I believe it is important to call a spade a spade. Otherwise, people get hurt. I'm willing to call something AI the moment it exhibits at least some properties that we associate with intelligence, like reasoning or abstraction. As far as I know, we haven't reached that level yet and I assume it will take us some time, especially because we still haven't figured out how brains actually work. But I'm no expert in AI or brain research.

@sophie@floss.social Hear, hear. I'm curious about AI for medical help. Could you share an example? I only hear and read about the dangerous stuff (like replacing doctors).