Related Links
19DeepSeek unveils next-gen AI model as Huawei vows ‘full support’ with new chips
DeepSeek has released its new V4 AI model, claiming cost-efficiency and competitiveness with top closed-source models from OpenAI and Google DeepMind. Huawei has pledged full support for DeepSeek with new chips.
Exclusive: Google deepens Thinking Machines Lab ties with new multibillion-dollar deal
Thinking Machines Lab, led by Mira Murati, has signed a multi-billion-dollar deal with Google Cloud for AI infrastructure using Nvidia's GB300 chips.
Subscribe to read
A company founded by former engineers from DeepMind and OpenAI has secured a $4 billion valuation after a deal involving Google's venture arm and Nvidia.
Google DeepMind
This article is about Google DeepMind, which is a company focused on artificial intelligence research and development.
Gemini Robotics ER 1.6: Enhanced Embodied Reasoning
Google DeepMind introduces Gemini Robotics ER 1.6, an enhanced embodied reasoning system for robotics.
Gemma 4
This entry refers to Google DeepMind's Gemma 4 model.
Gemma 4: Our most capable open models to date
Google announces Gemma 4, its most capable open models to date, which are purpose-built for advanced reasoning and agentic workflows.
AI Robotics Lab in Talks to Raise $1 Billion at $11 Billion Valuation
Physical Intelligence, a two-year-old robotics startup created by former Google DeepMind researchers, is in talks to raise $1 billion, which would bring the company's valuation to over $11 billion.
AI Science Startup Periodic Labs Is in Deal Talks at About $7 Billion Valuation
Periodic Labs, an AI startup founded by former OpenAI and DeepMind employees, is in talks to raise hundreds of millions of dollars at a $7 billion valuation.
Subscribe to read
A Financial Times article discusses Sebastian Mallaby's biography of Demis Hassabis, co-founder of DeepMind, and his pursuit of artificial general intelligence (AGI).
Measuring Progress Towards AGI: A Cognitive Framework
Google DeepMind is introducing a cognitive framework for AGI evaluation and launching a Kaggle hackathon to develop capability benchmarks. This aims to measure progress towards artificial general intelligence.
Spine Swarm Hits #1 on GAIA Level 3 and Google DeepMind DeepSearchQA
Spine Swarm achieved the #1 ranking on Google DeepMind's DeepSearchQA benchmark, surpassing the performance of AI models from OpenAI, Anthropic, and Gemini. The DeepSearchQA benchmark is known as one of the most challenging AI research benchmarks.
AI is rewiring how the world’s best Go players think
AI is transforming the world of Go, dominating training and influencing how top players approach the game. Ten years after AI's initial victory over a human Go champion, players are now grappling with the implications of AI-driven strategies on the traditional understanding of the game.
OpenAI and Google Workers File Amicus Brief in Support of Anthropic Against the US Government
Employees from OpenAI and Google, including Google DeepMind chief scientist Jeff Dean, have filed an amicus brief in support of Anthropic in its lawsuit against the US government. The brief expresses concerns that the government's actions could stifle AI innovation.
AlphaGo's next move
The DeepMind blog post discusses AlphaGo, a computer program that plays the board game Go. It alludes to advancements or new developments related to the AI program without providing concrete specifics.
Google Workers Seek ‘Red Lines’ on Military A.I., Echoing Anthropic
Over 100 Google AI employees have signed a letter addressed to chief scientist Jeff Dean, expressing their opposition to the use of the Gemini AI model in U.S. surveillance and certain autonomous weapons systems. This action echoes similar concerns previously voiced by employees at Anthropic, indicating a growing trend of AI researchers pushing for ethical boundaries in the technology's application.
SynthID
SynthID is a Google DeepMind tool for watermarking and identifying AI-generated content. It is designed to promote transparency and build trust in generative AI technologies.
AIs can generate near-verbatim copies of novels from training data
Researchers at Google DeepMind, leveraging datasets like BookCorpus, have demonstrated that large language models (LLMs) can memorize substantial portions of their training data, including near-verbatim copies of novels. This study highlights that the memorization capabilities of LLMs are more extensive than previously recognized, raising questions about copyright infringement.
Subscribe to read
OpenAI, Google DeepMind, and Anthropic are working on using advanced mathematics to better evaluate the true capabilities of AI models. These evaluations aim to provide more accurate insights into AI performance beyond traditional benchmarks.