Microsoft has struck hard by unveiling three new AI models focused on reasoning, named Phi-4: — The main model, Phi-4-reasoning (14B), outperforms OpenAI's o1-mini on benchmarks.
— The Phi-4-mini-reasoning (3.8B), despite its small size, rivals models twice as large on math
RESEARCH
-

Microsoft Unveils Phi-4 Reasoning AI Models Outperforming OpenAI o1-mini
By
–
-

Amazon Unveils Nova Premier Its Most Powerful AI Model Yet
By
–
Amazon a levé le voile sur Nova Premier, son modèle d’IA le plus puissant à ce jour : — Il est capable de gérer des tâches complexes, mais aussi de jouer le rôle de “professeur” pour affiner des modèles plus petits.
— C’est un modèle multimodal avec une capacité de contexte -
Google’s Gemini 2.5 Pro Completes Pokémon Blue Game
By
–
Google's Gemini 2.5 Pro just completed Pokémon Blue!
— Rowan Cheung (@rowancheung) 5 mai 2025
An independent software engineer live-streamed the game after taking inspiration from the (still unsuccessful) Claude Plays Pokémonpic.twitter.com/PFvQAebyEvGoogle's Gemini 2.5 Pro just completed Pokémon Blue! An independent software engineer live-streamed the game after taking inspiration from the (still unsuccessful) Claude Plays Pokémon
-

Llama-Nemotron: Open-Source Reasoning Models Match DeepSeek-R1
By
–
Meet Llama-Nemotron: a new family of open-source reasoning beasts with a twist—you can toggle between chat and deep reasoning on the fly. From 8B to 253B, they match top models like DeepSeek-R1 but run faster & leaner. And yes, it's all open for enterprise use. Tech
-

Study Warns AI Control Drops to 52% as AGI Approaches
By
–
Une nouvelle étude sur la "surveillance évolutive" de l'IA révèle des chiffres alarmants : même dans le meilleur des cas, nous ne contrôlons les IA avancées que 52% du temps, avec une efficacité qui diminue en approchant l'AGI. Max Tegmark estime à >90% la probabilité que la
-
OpenPhil contest prize awarded to AI risk skeptic with longer timeline
By
–
Eg: OpenPhil ran a "change our views" contest with money, which I didn't bother trying to submit to because I correctly predicted that they'd give the prize money to a "we need two Stalins" critic — in this case, somebody who argued for less AI risk and longer timelines.
-
LLM Chain of Thought as Educational Tool for Step-by-Step Problem Solving
By
–
It could be argued that when students see the chain of thought of LLMs they get examples of how to reason step by step to solve problems. So maybe that in itself could be valuable because students would be learning the value of step by step thinking and the different ways in
-
First Digital World War: Internet Security Battle Chronicle
By
–
First Digital World War The story of the battle between those who seek to exploit the Internet and those who work to protect it is chronicled in a gripping book by the author of "Black Hawk Down". The Conficker worm, which first infected a computer in November 2008, quickly
-
AI as Scientific Tool: Weather Modeling and Discovery
By
–
Thank you @eostudi0 for coming to @Caltech and interviewing me on #ai I talk about the need to keep being curious and use AI as a tool, rather than being afraid of AI. I talk about AI for scientific modeling and discovery, and training the first high-resolution AI-based weather
-

Python CheatSheets for Data Science and Machine Learning
By
–
CheatSheets! Giga Python Language. #BigData #Analytics #DataScience #AI #MachineLearning #IoT #IIoT #PyTorch #Python #RStats #TensorFlow #Java #JavaScript #ReactJS #CloudComputing #Serverless #DataScientist #Linux #Programming #Coding #100DaysofCode https://
geni.us/Py-Ch