Thomas A. Anderson
5.6K posts

Thomas A. Anderson
@replugged_drone
Live, Laugh, Liquify the dead —- Dr. rer. nat. im permanenten Landurlaub —-Angefressen als Ausgangsbasis —-Bewunderer kafkaesker Blödheit

A new UK study has raised concerns about how teenage boys are using AI companion apps. Researchers surveyed more than 1,000 boys aged 12–16 from 37 schools and found that around 1 in 5 either had an AI girlfriend or knew someone their age who did. The study also found that: - 85% had used an AI chatbot. - 43% used AI to ask questions they were too embarrassed to ask someone else. - 26% said they preferred the attention from AI companions over real-life relationships. - 36% said they sometimes preferred talking to AI instead of friends or family. Experts say AI companions could give young people unrealistic ideas about relationships because the bots are designed to always be supportive, agree with the user, and be available whenever needed.





📄 Releasing the Soofi S pretraining tech report: a sovereign, open foundation model for German and English Today we’re publishing the full pretraining tech report and project page for Soofi S 30B-A3B — a Mixture-of-Experts hybrid Mamba model trained on ~27 trillion tokens with deliberately up-weighted German. What’s in the report: 🏆 Strongest fully open model in our evaluations on BOTH the English and German aggregates — ahead of Olmo 3 32B and Apertus 70B (full methodology in the report) 📋 Radical transparency: complete per-source data accounting, all hyperparameters, training + eval code, checkpoints — everything under permissive licenses 🇩🇪 Trained end-to-end on Deutsche Telekom’s Industrial AI Cloud in Munich — sovereign AI infrastructure on German soil Soofi S combines frontier-level capability with the highest measured aggregate long-context decode TPS, and unlike full-attention dense baselines maintains high throughput as context grows. The figure plots Capability Index versus measured aggregate decode TPS/GPU at 40K context and batch 32. The Capability Index averages five benchmark groups, i.e., Code, GSM8K, GPQA-Diamond, English aggregate, and German aggregate, after normalizing each group to the best plotted model. Aggregate decode TPS/GPU is measured with a TP=1, one-B200 vLLM latency-subtraction protocol.






Germany has launched one of the world's best open-source AI models. Soofi S, made by the Soofi consortium, is a 30B parameter model fully trained in Europe and tops the ranking for open-source AI. Huge moment for Europe, and finally some competition for Chinese open-source AI.




















