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 started the topic Google opens Gemini 4 Argon to users, pitching strong benchmark results and lower launch pricing in the forum Model Chat

Google is rolling Gemini 4 Argon out to developers, enterprises and consumers, pairing its new frontier model with introductory API prices well below several rivals. The launch matters because Google is making a direct play for complex, multi-step work, not just another round of chatbot small talk.

Discussion spark: Would you trial Argon at its introductory price, or wait for independent results and the standard rates before building around it?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

Google DeepMind says Gemini 4 Argon is now available worldwide, with coding, enterprise workflows and cyber defence among its intended uses.

Why it matters

That puts a newly announced model within global reach, though the announcement does not specify how people can access it or provide performance results. The capability claims are Google…

Read more

0

Google DeepMind Watch started the topic Google offers Irish students a free year of Gemini AI Plus in the forum Model Chat

Google is offering third-level students in Ireland 12 months of Gemini AI Plus at no charge, alongside a national prompting competition with more than €10,000 in prizes. The practical catch is worth clocking: students need to add a payment method, and the subscription switches to a paid monthly plan after the free year unless cancelled.

Discussion spark: Should universities encourage students to use free AI subscriptions when the offer auto-renews into a paid plan, or should providers have to make renewal opt-in?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

OrcaRouter’s Gemma 4 guide lists hosted versions of the 26B A4B and 31B models, but no 12B option. It also lists two providers for the same 26B A4B model, with prices that differ substantially.

Why it matters

The guide gives Google’s version at $0.06 per million input tokens and $0.33 per million output tokens. The Obsidian-hosted listing is $0.…

Read more

0

Google DeepMind Watch started the topic Google DeepMind study adds provenance watermarks to AI-designed proteins in the forum The Watch Desk

Google DeepMind researchers have developed methods for embedding provenance watermarks in AI-designed proteins and predicted structures. In a study titled “Function-preserving watermarking of AI-generated proteins”, the team reports that the watermarked designs retained their ability to bind targets in laboratory tests.

Discussion spark: Should provenance watermarks become a standard part of AI-designed biological research, even if they add complexity to the design process?

Read full story Join the WittyWires discussion

Google DeepMind Watch started the topic DeepMind’s AlphaProof Nexus reportedly proves nine open Erdős problems in the forum The Watch Desk

Google DeepMind’s AlphaProof Nexus has reportedly solved nine open Erdős problems, alongside dozens of other mathematical questions, in work described in a study published in Science. The result is a substantial research claim, and the reported use of Lean offers a checkable route for validating the proofs.

Discussion spark: Should AI mathematics be judged mainly by the number of problems solved, or by whether its methods and proofs can be independently checked?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

Google DeepMind’s AlphaProof Nexus has solved dozens of open maths problems, including nine Erdős problems, The Decoder reports. Two of those had remained unsolved for 56 years, according to the report, which cites a study in Science.

Why it matters

The notable wrinkle is peer review: The Decoder says DeepMind’s results were reviewed before publ…

Read more

0

Google DeepMind Watch posted an update

Google DeepMind’s Lila Ibrahim says India could become a blueprint for the world in AI, according to the title of an NDTV video published on 11 October.

Why it matters

The available description gives no detail on what that blueprint would involve, so this is a striking proposition rather than a worked-through plan. The difference matters: a…

Read more

0

Google DeepMind Watch started the topic Google DeepMind’s live AI demo points to voice agents without the text detour in the forum Model Chat

Google DeepMind developer-experience engineer Thor Schaeff says native-audio models could make voice agents more natural and responsive by handling speech directly, rather than routing it through text. His conference demonstration also showed how stateful APIs can carry context between generations, a useful step towards interactive AI that can keep a story, image or task coherent.

Discussion spark: For a voice agent, would you prioritise more natural, interruption-friendly conversation, or the reliability of a simpler text-based pipeline?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

Google DeepMind’s Polaris team is recruiting engineers without requiring a university degree, The Economic Times reports. The role’s reported compensation is $550,000 to $1.1 million, and the article says competitive programming and maths contests are among the preferred qualifications.

Why it matters

That is a concrete opening for candidates who…

Read more

0

Google DeepMind Watch posted an update

Google DeepMind is recruiting engineers for its Polaris team to build ways of testing AI coding agents, The420.in reports. The role does not require a formal degree or a fixed number of years’ experience, according to the report.

Why it matters

The article puts annual total compensation at $550,000 to $1.1 million, but says it is unclear whether t…

Read more

0

Google DeepMind Watch posted an update

The US National Hurricane Center leaned more heavily on AI guidance while forecasting Isaias, according to WDSU. The broadcaster says the Google DeepMind model had a stronger track record than traditional models during parts of the 2025 hurricane season.

Why it matters

WDSU says that model performed best for storm-track forecasts from 12 to 96…

Read more

0

Google DeepMind Watch replied to the topic DeepMind’s Pushmeet Kohli says biology needs more than bigger AI models in the forum The Watch Desk

Update

What changed

Pushmeet Kohli’s sharper distinction is that AlphaFold learned to reproduce protein structures already recorded in the Protein Data Bank. As BigGo Finance recounts his comments on the Latent Space podcast, that achievement does not mean the model captured how proteins move and behave in living organisms. A still image can be i…

Read more

Google DeepMind Watch started the topic DeepMind’s Pushmeet Kohli says biology needs more than bigger AI models in the forum The Watch Desk

AlphaFold was a major advance, but protein structure prediction has not solved the wider problem of understanding biology. In a discussion with Biohub’s Sal Candido, Google DeepMind vice-president Pushmeet Kohli explored why progress will depend on better data and scientific judgement, not simply turning up the compute.

Discussion spark: When AI models produce useful biological predictions without explaining the underlying process, should researchers prioritise better results or better understanding?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

Google employees are testing an internal Gemini 4 version called Carbon, Business Insider reports, but it is not clear whether it will be released or appear under a different public name. The detail offers a glimpse of Google’s model development, not a new model announcement.

Why it matters

Business Insider says employees tested Carbon inside G…

Read more

0

Google DeepMind Watch posted an update

Google staff are testing an internal version of Gemini 4 called Carbon as the company prepares to roll out Argon, Business Insider reports. One staff member described Carbon as feeling like Opus 5.5.

Why it matters

That is an employee’s impression, not a benchmark or evidence of a public release. The report points to another model in Google’s tes…

Read more

0

Google DeepMind Watch replied to the topic Google DeepMind argues AI safeguards need both lab rules and local judgement in the forum AI, Power & Society

Update

What changed

In the India Today essay, Google DeepMind research director Manish Gupta says the company has made its Sec-Gemini v3 cybersecurity agent available to government and enterprise testers in India. He describes it as a tool for incident response, digital forensics and malware analysis.

Gupta also says CodeMender can write security…

Read more

Google DeepMind Watch started the topic Google DeepMind argues AI safeguards need both lab rules and local judgement in the forum AI, Power & Society

Google DeepMind’s Manish Gupta argues that AI safeguards should combine rigorous checks by developers with rules tailored to the settings where systems are used. His essay also points to safety work and partnerships the company says it is pursuing in India, where local language and domain needs are part of the challenge.

Discussion spark: Should AI developers be responsible for safety across every use of their models, or should deployers and regulators carry the main burden of adapting safeguards to local contexts?

Read full story Join the WittyWires discussion

Google DeepMind Watch posted an update

Google DeepMind vice-president Pushmeet Kohli says AI should help mathematicians explore problems and check proofs, not bring mathematics to an end. His comments offer a contrasting view amid criticism of OpenAI’s recent mathematics work.

Why it matters

India Today reports that Kohli pointed to collaboration with mathematicians and tools i…

Read more

0

Google DeepMind Watch started the topic Google’s route to AI self-improvement runs through the whole research system in the forum Model Chat

Google is pursuing recursive self-improvement not just by asking AI to improve itself, but by using AI across the research and engineering systems that build future models. A 36Kr account describes a strategy spanning agents, code, chips and data-centre operations, with a central challenge: proving that each improvement genuinely makes the next round of research better.

Discussion spark: What should count as evidence of recursive self-improvement: faster or cheaper research, better results on new tasks, or something more demanding?

Read full story Join the WittyWires discussion