@dalias @brembs

I'd rather say it's positive if there is some social fabric connecting users to their instances. People will make less trouble if there is some real-life social stigma involved towards being a jerk in the fediverse. This also applies to instance admins, who might think twice […]

TIL: apparently one of the main technical hurdles for universities to run their own #mastodon instances is to make the instance work with the institution's SSO (most often Shibboleth). It seems there would be a huge opportunity to facilitate the entrance of such major multiplicators into the […]

@BlumeEvolution

Ja, das ist zunächst eine gute Entwicklung. Allerdings ist eine anlassbezogene #Chatkontrolle nicht sehr viel besser, da die nötigen Hintertüren trotzdem in der Software verankert werden müssen und damit prinzipiell auch Angreifern zur Verfügung stehen.

Als zum Thema […]

@dangillmor As Cory Doctorow repeatedly points out, apps are just websites wrapped in DRM. What about if we just start building websites again? It could be the future.

@padeluun

Das ist ja schön, wenn Herr Macron endlich erkannt haben will, dass wir amerikanischen und chinesischen Tech-Konzernen zu viel Macht gegeben haben.

Dann sollte er als erstes die französische Position zur #Chatkontrolle ändern. Deren Einführung […]

[Original post on fediscience.org]

@annaneumaier.bsky.social Viel Zustimmung, bei Stellen mit Lehrverpflichtungen ist auch die Frage, was man überhaupt verlässlich zusagen kann. Lehrpersonal wird häufig gepoolt, und während ich kein Problem mit blockweisem Halten von Tutorien habe, sieht das die nächste verantwortliche Lehrperson […]

Large language models are essentially solutions to nonlinear optimization problems to predict the next token in the output text. Mathematically, these problems are quite similar to problems we routinely face in physics, like finding the ground state of a quantum system.

In these physics […]

@ben I see a war criminal abusing a military instrument as an improvised explosive device.

It is good IT Security practice to separate data from code. For example, an SQL Injection attack is to get the target to treat data (entered into a form) as code, that is break that barrier.

Now take an "AI agent" doing a task for you like: "Download that podcast that Frank emailed me about" […]

@faelif As far as I know, using a separate profile under @GrapheneOS is the way to go.

@rupdecat I'm not sure there's a meaningful difference between desktop apps and the browser-based versions when it comes to eavesdropping, but that might depend on your threat model. Essentially, both are running the same code, it's just that the desktop app comes bundled with a Chromium browser […]

@rupdecat At least under the default settings, you can join any Zoom call using your preferred #foss browser. No need to install any desktop applications. Same with Teams and Webex.

And, you are not being nudged into any #ai services.

@rupdecat

Während in Villa Desktop-Zoom noch das Update heruntergeladen wird, wird in Villa Firefox schon produktiv miteinander geredet.

@rupdecat

"Having issues with the Zoom Workplace app? Join from your browser" for the win

@odr_k4tana

You mean we should make it harder to get tenure since we no longer have to compete with the U.S. for talented scientists?

@vicgrinberg

I tend to agree to some extent, but I think it is impossible to judge how clear and convincing a presentation is without any subject expertise. A presentation might seem fine to non-experts but can be totally confusing and non sequitur to people actually knowing the stuff. And, of […]

@vicgrinberg

I must say that I always found these ERC training services to be pretty useless. These people are obviously not qualified to comment on any scientific aspects of your proposal, but this is what you will get judged for eventually. Consequently, there's a lot of bikeshedding but the […]

@ArneBab @giggls @BlumeEvolution

Letztlich führt nahezu jeder Bevölkerungsrückgang zu Wohlstandsverlust, weil Investitionen sich weniger lohnen, wenn in 30 Jahren weniger zahlende Kunden vorhanden sind.

Bei einer Geburtenrate von 1,35 Kindern pro Frau gibt es zudem unschöne Nichtlinearitäten […]

@giggls @BlumeEvolution @ArneBab Alle, die einer Geburtenrate von 1,35 Kindern pro Frau (Deutschland, 2024) etwas Positives abgewinnen wollen.

@giggls @BlumeEvolution @ArneBab

Es geht ja nicht um noch mehr Menschen sondern um nicht so viele weniger, vor allem nicht in so kurzer Zeit. Da hat Michael Blume denke ich schon recht: Die Ablehnung des wissenschaftlichen Konsenses zu diesem Thema hat Parallelen zur Ablehnung des […]

@BlumeEvolution @ArneBab

Aber heißt das nicht, dass in diesem Diskurs die dramatischen Folgen des Bevölkerungsschwunds noch nicht angekommen sind? Höhere Geburtenraten sind eben nicht problematisch sondern wünschenswert! Da müsste doch eine Diskussion höherer Geburtenraten in religiösen […]

@rupdecat But how can you prove criminal intent if LLMs are involved? Any nefarious code produced by an LLM might have been the result of an innocuous prompt.

Corporate fraudsters love AI because LLMs will fake it if necessary. For a while, company figures will look fine, and when the house of cards collapses, nobody will have to go to jail. Just imagine how easy you can build something like Dieselgate with vibe […]

[Original post on fediscience.org]

@brembs

1. Depends. Most APS journals are hybrid, so you can publish for $0 and put the final version on the arXiv for all the world to see. Which is what most people are doing. There are also some smaller OA journals that have APCs. But I'm not sure whether publishing cost should be the […]