Google DeepMind
Artificial intelligence research laboratory
Follow Google DeepMind on Notably News to receive short updates to your email — rarely!
| May 2025 | Google DeepMind unveiled AlphaEvolve, an evolutionary coding agent that uses Large Language Models like Gemini to design and optimize algorithms. The system can generate variations of algorithms, combine them, and select the best candidates through iterative processes. |
| May 2025 | Google announced Veo 3, which can generate both video and audio. |
| April 2025 | Google announced Veo 2 is now available in Gemini App for Advanced Users and Gemini Artificial Intelligence Studio, free for all users. |
| April 2025 | Google introduced DolphinGemma, a research AI model designed to potentially decode dolphin communication by learning the structure of dolphin vocalizations and generating novel dolphin-like sound sequences. |
| April 2025 | Google DeepMind announced Lyria, a text-to-music model, and Chirp 3, a voice-cloning feature capable of synthesizing speech in approximately 35 languages, at the Google Cloud Next Conference. |
| March 2025 | Google introduced TxGemma, an open-source model designed to improve efficiency in therapeutics development. |
| March 25 2025 | Google released Gemini 2.5, a reasoning model that pauses to 'think' before responding. Google announced that all future models will incorporate reasoning capabilities. |
| March 2025 | Google released Gemma 3, described as the most capable model that can run on a single GPU, available in four sizes: 1B, 4B, 12B, and 27B parameters. |
| March 2025 | DeepMind launched two AI robotics models: Gemini Robotics and Gemini Robotics-ER, designed to enhance robotic interaction with the physical world. |
| February 2025 | DeepMind launched PaliGemma 2 Mix, a multi-task fine-tuned vision-language model available in 3B, 10B, and 28B parameter versions with 224px and 448px resolutions. |
| 2024 | At the International Mathematical Olympiad, AlphaProof and an adapted version of AlphaGeometry achieved performance equivalent to a silver medalist in problem-solving across combined categories, marking a significant milestone in AI mathematical problem-solving capabilities. |
| 2024 | Google DeepMind published results of an experiment using large language models to identify areas of overlap among online group members, named the 'Habermas machine' project. In this experiment, AI-generated summaries were rated higher than human moderator summaries 56% of the time. |
| December 2024 | DeepMind released Genie 2, an advanced version of their generative AI model that expanded capabilities to create diverse and interactive 3D environments. |
| December 2024 | Google released Veo2 through VideoFX, featuring 4K resolution video generation and improved physics understanding. |
| December 2024 | Google introduced PaliGemma 2, an upgraded vision-language model expanding the Gemma model capabilities. |
| December 12 2024 | Google released Gemini 2.0 Flash, the first model in the Gemini 2.0 series. This version expanded multimodal capabilities to include image and audio generation, as part of Google's strategy to integrate advanced AI into autonomous agents. |
| October 2024 | Demis Hassabis and John Jumper received half of the 2024 Nobel Prize in Chemistry for protein structure prediction, specifically citing the AlphaFold2 achievement. |
| June 2024 | Google began releasing Gemma 2 models, expanding the Gemma language model series. |
| May 2024 | Google DeepMind announced Veo, a multimodal video generation model capable of creating 1080p videos over one minute long at Google I/O 2024. |
| May 2024 | Released AlphaFold3, which advanced protein structure prediction by improving benchmark test accuracies for interactions between proteins and various molecules. |
| March 2024 | DeepMind introduced SIMA (Scalable Instructable Multiword Agent), an AI agent capable of understanding and following natural language instructions across various 3D virtual environments, trained on nine video games from eight studios and four research environments. |
| March 2024 | Mustafa Suleyman was appointed as the EVP and CEO of Microsoft's newly created consumer AI unit, Microsoft AI. |
| February 21 2024 | DeepMind released the first Gemma large language models, including a 7 billion parameter GPU/TPU-optimized model and a 2 billion parameter CPU/on-device model, trained on up to 6 trillion tokens of text. |
| December 6 2023 | Google DeepMind released Gemini, a multimodal large language model that succeeds LaMDA and PaLM 2, designed to compete with OpenAI's GPT-4. Gemini was launched in three sizes: Nano, Pro, and Ultra. |
| November 2023 | Google DeepMind announced the Open Source Graph Network for Materials Exploration (GNoME), a tool that proposed millions of previously unknown materials, including 736 stable crystalline structures experimentally produced by MIT. |
| June 2023 | DeepMind announced AlphaDev, an AI system that discovered more efficient sorting and hashing algorithms using reinforcement learning. The new sorting algorithm was 70% faster for shorter sequences and 1.7% faster for longer sequences, and was accepted into the C++ Standard Library - the first algorithmic change in over a decade. The hashing algorithm was released to an open-source library. |
| June 2023 | DeepMind released RoboCat, an advanced AI model capable of controlling robotic arms and adapting to different robotic arm models and task types. |
| April 2023 | DeepMind merged with Google Brain division to officially become Google DeepMind. |
| March 2023 | DeepMind introduced 'Genie', an AI model capable of generating interactive virtual worlds from textual descriptions, images, or sketches, using an autoregressive latent diffusion model that allows frame-by-frame interactivity without labeled action data. |
| 2022 | Google Brain developed two text-to-image models called Imagen and Parti, creating a competitive offering to OpenAI's DALL-E image generation technology. |
| 2022 | Imagen development was transferred to Google DeepMind following the merger between Google Brain and DeepMind. |
| 2022 | Google DeepMind unveiled Ithaca, a deep neural network designed to restore damaged Greek documents, identify their date, and determine geographical origin. |
| 2022 | DeepMind unveiled AlphaCode, an AI-powered coding engine capable of creating computer programs at a performance level comparable to an average programmer. The system was tested on Codeforces programming challenges and achieved a ranking at the 54th percentile after being trained on GitHub data and Codeforces problems and solutions. |
| October 2022 | DeepMind released AlphaTensor, an AI system using reinforcement learning to discover novel matrix multiplication algorithms. The system found a more efficient algorithm for multiplying 4×4 matrices, reducing the number of distinct multiplications from 49 to 47, improving upon the Strassen algorithm that had been standard since 1969. |
| July 2022 | DeepMind announced DeepNash, a multi-agent reinforcement learning system capable of playing the board game Stratego at a human expert level. |
| July 2022 | DeepMind announced the release of over 200 million predicted protein structures, covering virtually all known proteins, in the AlphaFold database. |
| May 2022 | Released Gato, a polyvalent multimodal model trained on 604 tasks, capable of outperforming human experts on 450 tasks. |
| April 28 2022 | DeepMind published a blog post about Flamingo, a single visual language model (VLM) capable of accurately describing images with few training images. |
| March 2022 | Google fired AI researcher Satrajit Chatterjee after he questioned the findings of a Nature paper by Google AI team members Anna Goldie and Azalia Mirhoseini, which reported results from AI techniques for integrated circuit placement. |
| July 2021 | DeepMind released an open-source version of AlphaFold2 and completed protein structure predictions for nearly all human proteins and 20 other widely studied organisms, publishing these in the AlphaFold Protein Structure Database. |
| April 2021 | Samy Bengio left the Google Brain team, with Zoubin Ghahramani taking on his responsibilities. |
| February 2021 | Google fired Margaret Mitchell, a leader of the company's AI ethics team, allegedly for using automated tools to find support for Timnit Gebru. Following this, engineers began quitting in protest of Gebru's termination. |
| 2020 | Google Brain Team collaborated with University of Lille to develop the Interactive Speaker Recognition (ISR) model, an advanced automatic speaker recognition system that can identify speakers by requesting minimal user-specific words and enhance text-to-speech training security. |
| 2020 | Researchers from Google Brain, Intel AI Lab, and UC Berkeley created an AI model that enables robots to learn surgery-related tasks, such as suturing, by training with surgery videos. |
| 2020 | DeepMind introduced AlphaChip, a reinforcement learning neural architecture for chip placement that dramatically reduced chip design time from weeks to hours. The chip designs were subsequently used in every Tensor Processing Unit (TPU) iteration from this point forward. |
| 2020 | DeepMind presented WaveNetEQ, a packet loss concealment method based on WaveRNN architecture. |
| 2020 | DeepMind published Agent57, an AI Agent that surpasses human performance on all 57 games of the Atari 2600 suite. |
| 2020 | DeepMind made significant breakthroughs in protein folding with the AlphaFold system. |
| December 2020 | AI ethicist Timnit Gebru left Google after refusing to retract a paper on the potential risks of AI, including environmental impact, training data biases, and public deception. The request to retract the paper was made by Megan Kacholia, vice president of Google Brain. |
| January 2020 | Chris Lattner left the Google Brain team and joined SiFive. |
|
We are only showing the most recent entries for this topic. |
|
This contents of the box above is based on material from the Wikipedia articles Google DeepMind & Google Brain, which are released under the Creative Commons Attribution-ShareAlike 4.0 International License.