New Energy-Based Transformers AI Architecture Aims for Deeper ‘System 2’ Thinking

Markus Kasanmascheff / winbuzzer - AI researchers have unveil the Energy-Based Transformer (EBT), a new AI architecture for 'System 2' reasoning that promises better scaling but at a higher compute cost.The post New Energy-Based Transformers AI Architecture Aims for Deeper ‘System 2’ Think…

Back to Top / Friday, July 11, 2025, 12:20 pm / permalink 9879 / 2 stories in 7 months

Safe Superintelligence CEO Daniel Gross joins Meta’s new AI lab

Maria Deutscher / siliconangle - Meta Platforms Inc. has hired prominent artificial intelligence executive Daniel Gross and may reportedly buy a stake in his venture capital firm. Until last week, Gross (pictured) was the Chief Executive Officer of Safe Superintelligence Inc., a well-fun…

Back to Top / Thursday, July 3, 2025, 7:20 pm / permalink 9387 / 4 stories in 8 months

Google rolls out new Gemini model that can run on robots locally

Ivan Mehta / techcrunch - Google DeepMind has released a new language model called Gemini Robotics On-Device that can run tasks locally on robots without an internet connection.

Back to Top / Tuesday, June 24, 2025, 10:21 am / permalink 8699 / 5 stories in 8 months

Google Using YouTube Videos to Train Gemini, Veo 3 AI

Usman Qureshi / iphoneincanada - Google is training its Gemini and Veo 3 AI models using YouTube videos, raising questions about content consent, copyright, and AI model transparency.The post Google Using YouTube Videos to Train Gemini, Veo 3 AI first appeared on iPhone in Canada.

Back to Top / Thursday, June 19, 2025, 2:21 pm / permalink 8418 / 3 stories in 8 months

Apple Intelligence transcription is twice as fast as OpenAI's Whisper

appleinsider - Newly released to developers, Apple Intelligence's transcription tools are fast, accurate, and typically double the speed of OpenAI's longstanding equivalent.Apple Intelligence transcription tools being demonstrated by Apple — image credit: AppleThis coul…

Back to Top / Wednesday, June 18, 2025, 8:20 am / permalink 8302 / 4 stories in 8 months

MiniMax M1 model claims Chinese LLM crown from DeepSeek - plus it's true open-source

Thomas Claburn / theregister - China's 'little dragons' pose big challenge to US AI firms MiniMax, an AI firm based in Shanghai, has released an open-source reasoning model that challenges Chinese rival DeepSeek and US-based Anthropic, OpenAI, and Google in terms of performance and cos…

Back to Top / Tuesday, June 17, 2025, 2:20 pm / permalink 8244 / 2 stories in 8 months

Mistral Enters AI Reasoning Race with Magistral Model, But Benchmarks Reveal a Gap

Markus Kasanmascheff / winbuzzer - Mistral AI enters the AI reasoning race with its new Magistral models, offering a dual open-source and enterprise strategy that prioritizes speed and ecosystem over winning initial benchmarks against top rivals.The post Mistral Enters AI Reasoning Race wi…

Back to Top / Wednesday, June 11, 2025, 5:20 am / permalink 7624 / 4 stories in 8 months

o3-pro

Simon Willison / simonwillison - o3-proOpenAI released o3-pro today, which they describe as a "version of o3 with more compute for better responses".It's only available via the newer Responses API. I've added it to my llm-openai-plugin plugin which uses that new API, so you can try it ou…

Back to Top / Tuesday, June 10, 2025, 4:20 pm / permalink 7547 / 4 stories in 8 months

Mistral releases a pair of AI reasoning models

Kyle Wiggers / techcrunch - Mistral released Magistral, its first family of reasoning models. Like other reasoning models — e.g. OpenAI's o3 and Google's Gemini 2.5 Pro — Magistral works through problems step-by-step for improved consistency and reliability across topics such as mat…

Back to Top / Tuesday, June 10, 2025, 10:21 am / permalink 7463 / 3 stories in 8 months

Apple's new Foundation Models framework adds on-device AI to apps with three lines of Swift code

Matthias Bastian / the-decoder - Apple is rolling out new AI tools for developers, embedding generative models directly into Xcode and iOS apps with a strong emphasis on privacy and user control.The article Apple's new Foundation Models framework adds on-device AI to apps with three line…

Back to Top / Monday, June 9, 2025, 1:22 pm / permalink 7336 / 7 stories in 8 months

Researchers build massive AI training dataset using only openly licensed sources

Jonathan Kemper / the-decoder - The Common Pile is the first large-scale text dataset built entirely from openly licensed sources, offering an alternative to web data restricted by copyright.The article Researchers build massive AI training dataset using only openly licensed sources app…

Back to Top / Friday, June 6, 2025, 2:21 pm / permalink 7067 / 3 stories in 9 months

DeepSeek upgrades R1 AI model as it gears up for R2 release

DeepSeek has rolled out a minor but notable update to its R1 AI model, setting the stage for the forthcoming R2 release. The move, intended for trial by eager users, adds a polished layer of performance to an already competitive offering in the AI race—proving that even in tech, every little tweak counts.

Back to Top / Wednesday, May 28, 2025, 9:20 am / permalink 5925 / 2 stories in 9 months

Starburst Launches Major Platform Update for Enterprise AI

Starburst Data has unveiled substantial upgrades including advanced AI workflows and smarter data governance to ease enterprise AI implementations. The improvements are designed to alleviate data access bottlenecks and streamline analytics across distributed systems—a robust leap forward in making enterprise artificial intelligence more effective across industries.

Back to Top / Monday, May 19, 2025, 9:20 am / permalink 5011 / 2 stories in 9 months

AI Breakthrough in Decoding Animal Sounds Raises Communication Debates

Researchers employing advanced machine learning techniques are now decoding animal sounds—from dolphins to dogs—prompting lively debates about whether humans should try to “chat back.” The innovative study raises both scientific enthusiasm and ethical questions about interfacing with nature.

Back to Top / Saturday, May 17, 2025, 6:20 am / permalink 4944 / 1 stories in 9 months

DeepMind’s AlphaEvolve AI Shakes Up Code and Math Battles

Google DeepMind has unveiled its new AI agent, AlphaEvolve, which leverages Gemini-powered capabilities to write code, optimize algorithms, and tackle challenging math problems. The breakthrough is touted to save millions in computing costs while pushing the envelope of what AI can achieve—yes, even superheroes get jealous.

Back to Top / Wednesday, May 14, 2025, 12:20 pm / permalink 4700 / 7 stories in 9 months

Mistral Launches High-Performance Medium 3 AI Model

French startup Mistral has unveiled its latest AI offering, Medium 3, engineered to deliver industry-leading performance without breaking the bank. The dual reports highlight the model’s efficiency and competitive pricing, signaling a bold stride in the crowded AI market with a touch of irreverent French style.

Back to Top / Wednesday, May 7, 2025, 11:20 am / permalink 4231 / 6 stories in 9 months