Sunday, June 21, 2026 Trend Press · Cloudflare Pages

The Trend Tribune

"All the trends that are fit to read" Morning Edition Free of Charge
TODAY'S LEAD STORY

Nobel Prize-Winning Scientist Leaves DeepMind for Anthropic — The Era When AI Talent Wars Move 'Researchers'

John Jumper, who won the Nobel Prize in Chemistry for AlphaFold, has left DeepMind to join Anthropic. We examine the structural shifts in the AI industry revealed by the increasing mobility of top-tier researchers.

By tech_tanaka · June 21, 2026
Nobel Prize-Winning Scientist Leaves DeepMind for Anthropic — The Era When AI Talent Wars Move 'Researchers'
(Photo: illustrative)

What a Nobel Laureate's Move "From Google to a Competitor" Means

On June 20, 2026, a significant piece of news rippled through the AI industry. John Jumper — who helped create AlphaFold, the protein structure prediction model, and jointly received the 2024 Nobel Prize in Chemistry — was revealed to be leaving Google DeepMind to join Anthropic.

Jumper was a core researcher on the AlphaFold project, and his name is globally recognized in the context of AI and life sciences. His choice to leave Google less than two years after receiving the Nobel Prize has been interpreted by the industry as more than a simple career move — it reads as a "message."

According to TechCrunch's reporting, Jumper is just one of several "big names" departing DeepMind. Multiple prominent researchers are said to be leaving Google's AI division, painting a picture of a talent flow shifting away from Google and toward its rivals.

Why Anthropic?

Anthropic is a company founded in 2021 by Dario Amodei and others who originally held senior positions at OpenAI. It has positioned "AI safety" as its central research focus and develops the Claude series of large language models. Having recently secured massive investments from both Amazon and Google, it has rapidly grown in financial clout as well.

Several hypotheses exist for why a "titan of fundamental research" like Jumper would choose Anthropic: freedom of research, alignment with the organization's direction, or an attractive compensation package. While the specific reasons have not been made public, it seems likely that a shared language around AI safety is serving as a powerful draw for Anthropic.

The Industry Shift Revealed by Talent Mobility

Researcher relocations have become commonplace in the AI industry over the past few years. What makes this case stand out, however, is the symbolic weight of the words "Nobel laureate."

Researchers who once abandoned the stability of university tenure to enter industry are now moving between companies — a trend that suggests the "center of gravity" for AI research is no longer anchored to any single organization. Despite housing DeepMind, one of the world's premier research institutions, Google is finding it increasingly difficult to retain talent.

At the same time, this "battle for star researchers" casts a shadow of its own. If research continuity and team stability are undermined, long-term projects will inevitably suffer. The question of what happens to follow-on research to AlphaFold naturally comes to mind.

Other Voices on "Trustworthy AI"

On the same day as this news of the move, Meredith Whittaker, president of Signal, sounded a different kind of alarm about the AI industry. "AI chatbots are not your friends. They are not conscious beings, nor are they entities with feelings" — her remarks were a critique of the current trend of designing and marketing AI products as though they possess a personality.

On the talent side, leading researchers are gravitating toward more "safety-oriented" companies; on the product side, concerns are growing about designs that encourage users to form emotional dependencies. In 2026, the AI industry finds itself under unprecedented tension between technological advancement and ethical responsibility.

Conclusion

John Jumper's move to Anthropic is a symbolic event illustrating that the talent map of the AI industry is being redrawn. In an era when researchers get to choose where they work, a company's research philosophy and culture become its recruiting power. This news, which signals that Google's dominance in AI is far from unassailable, makes future talent movements well worth watching.

References: techcrunch.com / techcrunch.com
AIDeepMindAnthropic人材流動機械学習

The Robot 'Eye' Has Evolved — RealSense D585 Pro Ushers in a New Era of AI-Native Depth Cameras

By robo_takahashi

The RealSense D585 Pro features a proprietary Gen 5 SoC, delivering more than twice the depth quality of its predecessor. This highly anticipated new product, built with an AI-native design, fundamentally elevates the perceptual capabilities of robots.

Robot "Eyes" Have Come This Far

Anyone who has worked on robot development knows all too well how sensor performance sets the ceiling for the entire system. No matter how sophisticated a control algorithm you write, there's nothing you can do if the input data is poor. In that sense, the D585 Pro, announced by RealSense in June 2026, was genuinely exciting news for anyone in the robotics field.

What Makes the D585 Pro So Impressive

The D585 Pro's standout feature is its proprietary fifth-generation SoC (System-on-Chip), which RealSense developed in-house. Its spec sheet claims more than twice the depth quality of the previous generation.

"Depth quality" here isn't simply about resolution. It means a comprehensive improvement across factors that matter in real-world deployment — lower noise, better edge accuracy, improved tracking of moving objects, and stability in low-light environments. Anyone who has implemented bin picking for a robotic arm or obstacle avoidance for a humanoid robot knows firsthand just how much even a little depth camera noise can wreak havoc on a control system.

The other key concept is the "AI-native" design philosophy. With conventional depth cameras, the standard approach was a division of labor: the hardware outputs raw point cloud data, and AI processing is handled by an upstream computer. The D585 Pro pushes that pipeline into the camera itself. By running AI inference on the SoC, the design reduces the transfer load on the host while boosting real-time performance. Perhaps the clearest way to describe it is that edge computing and robot vision have begun to merge.

Why Now?

The backdrop to this announcement is the rapid expansion of the robotics market as a whole. According to a report from the International Federation of Robotics (IFR), the U.S. robotics industry recorded double-digit growth in 2025, driven largely by the food industry and non-manufacturing sectors.

The fact that robots are beginning to move outside of manufacturing lines means there is a growing number of cases where robots are being deployed in unstructured environments — the spaces where people normally live and work. Operating reliably in environments where lighting conditions and object placement change every time, rather than the fixed layouts of a factory floor, makes sensor precision and robustness absolutely critical. The timing of the D585 Pro's arrival aligns perfectly with this demand from the market.

What Stands Out from a Software Engineer's Perspective

As an engineer, what personally interests me is how the AI-native design will manifest at the SDK and API level. If AI processing is completed inside the camera, there is a possibility that "pre-interpreted data" arrives on the application side. If the filtering and segmentation of raw point clouds that developers previously handled themselves becomes unnecessary, development effort could be reduced substantially.

Of course, there is also the risk of black-boxing. The challenge of developers struggling to understand what AI models are running inside the camera — along with the biases and limitations of those models — is one that constantly haunts the world of physical AI. Particularly for safety-critical applications, there will likely be growing calls for transparency in this area.

In Closing — The Age of Smarter Sensors

Robot intelligence has, until now, largely been discussed in terms of software and control algorithms. In reality, however, what fundamentally determines the performance of an entire system is what the sensors see and how accurately they see it. The arrival of the D585 Pro signals that the direction of cameras themselves becoming "smarter" is finally entering full swing.

As the robotics market continues its double-digit growth and expands into unstructured environments, AI-native depth cameras have the potential to become a powerful tool for developers. If the opportunity arises to get hands on the hardware, I would love to put it through its paces in the kind of cluttered, real-world environment you find outside of a factory.

References: www.therobotreport.com / www.therobotreport.com / spectrum.ieee.org
ロボット深度カメラRealSenseフィジカルAIセンサー

AI to Save the UK Housing Shortage——The Full Picture of the 'Building Permit AI' Developed by DeepMind and the British Government

By thesis_yamamoto

The UK government has partnered with Google DeepMind to develop an AI prototype that accelerates the review process for housing construction permits, drawing attention as a potential solution to the chronic housing shortage.

AI Takes on the UK's "Housing Crisis"

The UK has long struggled with a severe housing shortage. One of the biggest barriers to keeping up with new home construction has been the complex and time-consuming planning permission process. In this procedure, local authority staff manually review vast quantities of application documents, cross-referencing them against regulations and assessing impacts on local residents — a process that commonly takes anywhere from several months to over a year.

In a bid to clear this chronic bottleneck, it has been revealed that the UK government and Google DeepMind have collaborated to develop an AI-powered planning permission review prototype.

AI Takes On the Task of "Swimming Through a Sea of Paperwork"

According to a blog post published by DeepMind in June 2026, the AI system rapidly analyses the enormous volume of documents involved in planning applications — including design drawings, environmental impact assessments, and land use regulation texts — to support the decision-making of planning officers.

The following capabilities are envisioned:

• Automatic document classification and summarisation: AI structures application materials running to hundreds of pages and extracts the key points that reviewers need to check • Regulatory compliance checks: Automatically cross-references local planning policies and national regulations, instantly flagging any apparent non-compliance • Similar case referencing: Searches past approval and rejection cases for comparable examples, supporting consistent decision-making

These features are expected to significantly shorten certain stages of the review process, allowing staff to focus their attention on complex considerations that require human judgement.

Why Did the UK Government Choose DeepMind Now?

The UK government currently has an ambitious target of building 300,000 new homes per year by 2029. Yet the construction industry is consistently vocal: "We can't build because we can't get permission." Local authority planning departments face ongoing staff shortages and budget cuts, leaving them without the capacity to handle growing application volumes.

The rationale for choosing DeepMind is clear. The company has a strong track record in science and engineering — with achievements such as AlphaFold and AlphaCode — and is renowned for developing AI that identifies optimal solutions within complex rule systems. The "rules-based complexity" of housing reviews maps closely onto exactly the kind of problem structure at which DeepMind excels.

New Challenges Raised by "Government × AI"

While technical expectations are running high, challenges are also coming into focus.

The first is the issue of transparency and accountability. A planning refusal is an administrative decision with direct implications for property rights; if humans cannot understand or explain the basis on which AI reached its conclusions, legal and political risks arise. The fact that this prototype stops at "supporting staff" rather than "automating decisions" appears to reflect precisely this concern.

The second issue concerns data quality and regional variation. Local planning policies in the UK differ from one authority to another, and in many areas historical case data has yet to be digitised. Since AI performance depends on data quality, establishing high-quality datasets is an essential prerequisite.

Finally, maintaining community involvement is also critical. The planning permission process is one of the few opportunities local communities have to engage with decisions affecting their environment, and there remains deep-seated concern that, in the name of efficiency, meaningful citizen participation could be reduced to a formality.

A Litmus Test for Administrative Efficiency AI

What makes this initiative so noteworthy is not only its technical novelty. It is attracting global attention as an example of a developed-world government making the decision to embed AI into a core public administration process.

If results are delivered in a domain as concrete and measurable as housing permissions, the door opens to expansion into other areas of public administration — tax assessment, environmental permits, public procurement, and beyond. Conversely, if problems emerge, trust in AI for government use could be seriously damaged.

As a prototype developed jointly by DeepMind and the UK government, this project stands as a litmus test for the era in which AI "solves social challenges beyond the laboratory" — and its progress will be watched with keen interest.

Source: deepmind.google
DeepMindAI行政住宅政策英国公共AI
Advertisement300 × 250