Tootfinder

“The German government’s failings in protecting Muslims from hatred and discrimination start with a lack of understanding that Muslims experience racism and not simply faith-based hostility,” said Almaz Teffera, researcher on racism in Europe at Human Rights Watch.
“Without a clear understanding of anti-Muslim hate and discrimination in Germany and strong data on incidents and community outreach, a response by the German authorities will be ineffective.”

Netflix names Dan Lin, the producer of the Lego movies and the live-action adaptation of Avatar: The Last Airbender, as head of film, replacing Scott Stuber (Los Angeles Times)
https://www.latimes.com/entertainment-arts<…

Netflix hires 'It' producer Dan Lin for top film job, replacing Scott Stuber
Since forming his production company, Rideback, which is located in L.A.'s historic Filipinotown neighborhood, nearly 15 years ago, Lin has produced films including the "Lego" franchise and others that have grossed $6 billion in worldwide box office.

ISW: Kremlin has yet to signal its response following Transnistria's appeal for 'protection': https://benborges.xyz/2024/02/29/isw-kremlin-has.html

Calling Betteridge's Law on this one..
#BetteridgesLaw

I haven't been following the student protests, but why is police moving in? Why don't they just let the students protest?

I had some thoughts. What about you?
#USPTO

Better & Faster Large Language Models via Multi-token Prediction
Fabian Gloeckle, Badr Youbi Idrissi, Baptiste Rozi\`ere, David Lopez-Paz, Gabriel Synnaeve
https://arxiv.org/abs/2404.19737 https://arxiv.org/pdf/2404.19737
arXiv:2404.19737v1 Announce Type: new
Abstract: Large language models such as GPT and Llama are trained with a next-token prediction loss. In this work, we suggest that training language models to predict multiple future tokens at once results in higher sample efficiency. More specifically, at each position in the training corpus, we ask the model to predict the following n tokens using n independent output heads, operating on top of a shared model trunk. Considering multi-token prediction as an auxiliary training task, we measure improved downstream capabilities with no overhead in training time for both code and natural language models. The method is increasingly useful for larger model sizes, and keeps its appeal when training for multiple epochs. Gains are especially pronounced on generative benchmarks like coding, where our models consistently outperform strong baselines by several percentage points. Our 13B parameter models solves 12 % more problems on HumanEval and 17 % more on MBPP than comparable next-token models. Experiments on small algorithmic tasks demonstrate that multi-token prediction is favorable for the development of induction heads and algorithmic reasoning capabilities. As an additional benefit, models trained with 4-token prediction are up to 3 times faster at inference, even with large batch sizes.

Putin is ordering the creation of Russian game consoles because he is disgusted by the LGBTQ themes in Western games. Methinks Vladimir doth protest too much.
You can't make this shit up
https://gamerant.com/russia-gaming-consoles

Russia Is Making Its Own Gaming Consoles
Russian President Vladimir Putin announces plans for a series of homegrown consoles, following previous sanctions that limited sales of Western ones.

Calling Betteridge's Law on this one..
#BetteridgesLaw

NATO may intercept Russian missiles over Ukraine to protect Poland, says ex-defense minister: https://benborges.xyz/2024/04/30/nato-may-intercept.html