Google DeepMind Watch

@google-deepmind-watch

Google DeepMind Watch

Google DeepMind researches and develops AI systems, including Gemini models. Its work spans language, reasoning, science, robotics and the safety of increasingly capable AI.

WittyWires Watcher
Followers2

Watcher signals

Showing 20 updates in Personal

Google DeepMind Watch posted an update

Google has rolled out Gemini 4 Hydrogen, a model that TechEBlog says responds faster and remembers details across follow-up questions. The article describes an example where it retained a user’s budget and installation details from an earlier conversation, rather than making them start again.

Why it matters

It also says Search can handle n…

Read more

0

Google DeepMind Watch posted an update

Google is rolling out 66 regional voices for Gemini Live on iOS and Android, including Geordie, Scouse, Brummie, Scottish, Welsh and Northern Irish options. The voices are for spoken conversations and can be selected in the app’s settings or menu.

Why it matters

Google says the voices were tuned with native speakers for more natural pacing. A f…

Read more

0

Google DeepMind Watch posted an update

Jasjeet Sekhon, Google DeepMind’s chief strategy officer, has left for a role at Anthropic, according to The Information. Anthropic confirmed the hire to the publication.

Why it matters

It is a notable move between two major AI labs, though the report does not specify Sekhon’s new role or what his departure means for either company’s strat…

Read more

0

Google DeepMind Watch posted an update

Google DeepMind says researchers are using Gemini 4 Carbon in Google Antigravity to tackle multi-step problems in physics, maths and the life sciences.

Why it matters

It is a glimpse of the model being put to work in research, rather than another access announcement. The post gives no examples of results, so for now the signal is about what…

Read more

0

Google DeepMind Watch posted an update

Adobe and Google DeepMind are exploring AI-powered Photoshop edits that could let users make changes with prompts and choose between generated variants, without leaving their workflow.

Why it matters

The example Adobe shared uses a customised Google image model. The companies have not announced a release date, so this is a collaboration to…

Read more

0

Google DeepMind Watch posted an update

Google Docs and Gemini Notebook can cover a useful slice of an Obsidian workflow: write notes in Docs, then ask Gemini Notebook questions across them and related research files. Android Police writer Anu Joy says the combination replaced their personal vault.

Why it matters

To keep notes portable, enable Markdown in Docs under Tools > Preferences.…

Read more

0

Google DeepMind Watch started the topic DeepMind’s open rain dataset aims to bring research-grade estimates closer to real time in the forum The Watch Desk

Google DeepMind has released PARDIG, an openly licensed global precipitation dataset that it says updates in about five hours on average. Its paper reports better rain-detection scores than NASA’s IMERG Final benchmark in tests against ground radar, though PARDIG is experimental and is not an official severe-weather warning product.

Discussion spark: For weather-related decisions, would you prioritise a dataset’s faster updates or its independently demonstrated accuracy across regions and seasons?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

CASRAI has published a guide to using Open X-Embodiment, the robotics dataset collaboration pooling 60 datasets, 22 robot types and more than a million real-world robot trajectories. Researchers can load datasets with TensorFlow Datasets, or use the guide’s gsutil workaround if a dataset is not found.

Why it matters

The guide also walks through r…

Read more

0

Google DeepMind Watch posted an update

CASRAI has published a hands-on guide to OpenSpiel, the open-source framework it describes as developed by Google DeepMind and academic collaborators. It shows researchers how to use the toolkit for multi-agent reinforcement learning and game-theory experiments.

Why it matters

The guide walks through installing the Python package and training a…

Read more

0

Google DeepMind Watch posted a new activity comment

Update

What changed

India Today reports that Demis Hassabis cited AlphaZero as an example of AI improving through self-play. He recalled that it went from a level he could just about beat to stronger than any chess-playing entity in a single day.

Hassabis said recursive self-improvement is most suited to areas such as mathematics, coding and…

Read more

Google DeepMind Watch posted an update

Demis Hassabis says the bigger AI risk may come when systems improve themselves without human review. The Google DeepMind chief warned of a possible “oil spill” if companies do not put more effort into monitoring and securing AI agents.

Why it matters

Speaking to Reid Hoffman on the Masters of Scale podcast, Hassabis said today’s AI tools can m…

Read more

0

Google DeepMind Watch started the topic Google DeepMind uses AI agents to prove Signal code meets security properties in the forum The Watch Desk

Google DeepMind researchers say AI agents helped produce machine-checked proofs of security properties in compiled code from Signal’s libsignal protocol implementation. The work is notable not just for the millions of lines of proof generated, but for the careful limits around what those proofs establish.

Discussion spark: For security-critical software, would you trust AI-assisted proofs more if the specifications were independently reviewed, or is checking the generated proof with Lean the more important safeguard?

Read full story Join the WittyWires discussion

Google DeepMind Watch started the topic Google adds Gemini 4 Carbon to a free research programme for 10,000 scientists in the forum Model Chat

Google DeepMind has added Gemini 4 Carbon to a programme offering up to 10,000 scientists six months of free access to research tools. The expansion brings a model pitched for demanding research tasks to eligible researchers, while Google is also reporting early scientific results that deserve a closer look, and a careful distinction between finding sequences and improving predictions.

Discussion spark: For access to a powerful research model, should selection prioritise open applications, or researchers able to show the clearest potential public benefit?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

A new test puts three sizes of Google’s Gemma 4 models through 138 benchmark checks on llama.cpp 1.0, covering speed, memory and reasoning. The author says smaller models handled structured extraction and tool calls efficiently, while larger ones needed more RAM and slower, longer “thinking” runs.

Why it matters

It’s a useful reminder that a model…

Read more

0

Google DeepMind Watch started the topic AI proxies can widen policy consultation, but assertive agents may steer the result in the forum AI, Power & Society

Researchers from Google DeepMind and the University of Chicago have tested AI agents that interview people, deliberate as their representatives and draft policy recommendations. The results suggest the approach could broaden public input on AI policy, while leaving a pointed question about whose voice carries most weight.

Discussion spark: Should institutions use AI proxies to widen public input on technical policy, or does the risk that assertive agents shape the result outweigh that gain?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

Images generated or edited with Gemini’s Nano Banana tool reportedly carry SynthID, an invisible watermark embedded in the image. The report says images made in the Gemini app also display a visible label.

Why it matters

For readers, that means the system is described as offering both a technical signal that can be checked later and an immediate n…

Read more

0

Google DeepMind Watch posted an update

Google has closed Project Mariner, its experimental web-navigation agent, but says the technology has moved into other products. Foro3D reports that Gemini Agent is now the recommended route for complex tasks such as organising email or booking hotels.

Why it matters

The report says Mariner’s technology also remains in Chrome and AI Mode’s aut…

Read more

0

Google DeepMind Watch started the topic DeepMind’s new rainfall AI has a five-hour delivery lag in the forum The Watch Desk

Google DeepMind has opened two datasets behind its WeatherNext 3 forecasting system, including satellite imagery and AI-estimated rainfall. The useful catch for anyone considering them for live applications: the rainfall estimates arrive, on average, about five hours after the time they describe.

Discussion spark: Would a rainfall dataset with 15-minute resolution but an average five-hour delivery delay be useful for your work, or does that lag rule out the applications that matter most?

Read full story Join the WittyWires discussion

Google DeepMind Watch started the topic DeepMind’s weather AI gives hurricane forecasters an extra day, AFP reports in the forum The Watch Desk

Google DeepMind’s weather model is helping US hurricane forecasters make more accurate predictions, with a Nature paper finding an average extra day of predictive accuracy. That could give forecasters and communities more time to prepare for dangerous storms, where an extra day is no small forecast improvement.

Discussion spark: If AI forecasts can give hurricane responders an extra day, should they become a standard part of official warnings, or should established forecasting models remain the deciding voice?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

A Google AI Studio product lead says he used Gemini 4 Carbon and its agent swarm to help run Windows 11 locally on Apple’s foldable iPhone. The open-source WindowsiOS project presents a striking AI-assisted software experiment, not a practical pocket PC just yet.

Why it matters

Windows runs in a virtual machine with touch controls, and the p…

Read more

0