> Um die Klimaziele zu erreichen, soll der Bahnverkehr in Deutschland eigentlich ausgebaut werden. Doch nun droht sogar, dass künftig weniger Regionalbahnen fahren. Denn den Ländern geht offenbar das Geld aus. Von B. Hümbs.
Pro-Tipp: Wer sichergehen will, das Gendern-Verbot in Bayern korrekt umzusetzen, und die Frauen in der Sprache wieder unsichtbar zu machen, benutzt dort am besten nur noch männliche Schreibformen: Sehr geehrte Herren, Krankenbruder statt Krankenschwester, Marius statt Maria usw. #bayern#GenderVerbot
> Bei der "Neuen Zürcher Zeitung" herrscht offenbar Aufregung: Man hört von angeblichem "Rechtsruck", von "interner Zensur" und "Säuberungswellen". Was ist dran am Wirbel?
Sollte sich das bestätigen, ist das beschämend und räumt die Berliner Bildungssenatorin hoffentlich freiwillig ihre Stuhl. #berlin#bildungspolitik#Günther-Wünsch #CDU
> I'm sleep-deprived, completely exhausted, but incredibly excited about the whole experience. Here are my notes, impressions and a lot of pictures about FOSDEM 2024.
#IPCEI-CIS will get more than 2 billion Euro to develop a "Multi-Provider Cloud -Edge Continuum" as announced on #fosdem2024. The European Commission approved the money beginning of December 2023. The aim is to develop the next-generation cloud infrastructure and services.
Agnès Crepet and Luca Weiss talk about open source for #Fairphone now. Support for the proprietary chipsets from Qualcomm in Fairphone is end-of-life after a few years. Then Fairphone has to take over, for a longer support.
Fortunately, they support #opensource software like #lineageOS and #PostmarketOS. Also, better support in #Linux for using new kernel versions - though not feature complete yet. #TWRP builds for Fairphone available, plus support for many ROMs.
For building an #AI#pipeline, Tuana Çelik proposes to take a look at the Python-based #Haystack framework. It looks promising and consists of so-called components that can be arranged into pipelines. There are demo projects available on HuggingFace. #fosdem2024
Martin Hickey talks about using APIs with foundation models using frameworks like HuggingFace, Ray and Triton Inference Server. Basically, AI can be used as a normal function in code now. #fosdem#LLM#ai
Wouldn't it be nice if we could explain AI reasoning? Dynamic Cause Modelling #DCM is one path. It tries to make predictions on dynamic data. The speaker is working on making Mathlab code available in C++. https://github.com/embecosm/dcEmb
Now again #LLMs: if you do not want to train your own #ai foundation model, you can patch it with so-called #adapters. Benjamin Trim talked about their own open-source adapter micro framework: #Refiners work on top of #PyTorch and use declarative layers to patch models, context API to store state. #fosdem2024
Demo time: WiFi is not working, changing to video. Demo gods still in a bad mood at #fosdem2024
There are a lot of good videos on YouTube about doing RAG with Langchain, also with open-soure models. #Langchain#ai#python#fosdem2024
Stefano Fancello is talking about #Langchain, an open-source Python-based toolkit for Retrieval Augmentation Generation #RAG. It helps preparing your own data as a context for a question you send to a Large Language Model #LLM. Langchain tools can ingest all kinds of document formats, split documents into Chunks, and create so called #Embeddings and send it to the LLM. #fosdem2024#ai#opensource
Problem: Models are black boxes and results not reproducible and auditable from pure usage. #AGI#fosdem2024 Government oversight should be there. Nadella from Microsoft wants AI for pure market growth. OpenAI works on Superalignment with 20 percent of resources. Speaker thinks: Safety should be 80 percent! And we need a debate in #FOSS about it.
Reminder of the potential range of risks connected to #AGI appearance. Safety should be priority, therefore reproducibility and a scientific approach is important. Needs availability of code and data (not the OpenAI approach anymore, I might add) #fosdem2024#opensource