Google DeepMind engineer Robert O'Callahan resigns over rapid AI progress, warning of uncertain ASI risks while rejecting claims of certain AI doom.

Introduction

A Google DeepMind engineer has resigned from what he described as an “amazing” job, citing concerns over the speed at which artificial intelligence is advancing. Robert O'Callahan, an Auckland-based engineer, said in a resignation email published on his blog that AI progress is moving “far too rapid” and that his own work was contributing to making advanced AI systems cheaper and faster.

However, O'Callahan's position does not align completely with the most extreme warnings about artificial intelligence. While he considers the possibility of artificial superintelligence, or ASI, causing catastrophic harm to be real, he said he is not convinced that AI doom is inevitable.

His decision highlights the growing debate inside the technology industry over whether the rapid development of increasingly capable AI systems is moving faster than society's ability to understand and regulate the risks.

Why Did Robert O'Callahan Leave Google DeepMind?

O'Callahan worked on hardware chip-design tools rather than directly developing AI models. Initially, he considered his work relatively harmless because he was not building the AI systems themselves.

Over time, however, he concluded that his work could have a significant indirect impact on AI development.

According to his account, the tools he worked on would help accelerate the design of a new generation of AI chips. Those chips could ultimately make AI systems faster, cheaper and more widely available.

That realization changed how he viewed his role.

He said his team's objective of making AI “much cheaper and lower-latency” was not necessarily beneficial for society at the present moment.

Engineer Warns About ASI Risk Without Predicting Certain Doom

O'Callahan's position is notable because he does not claim that artificial intelligence will inevitably cause human extinction.

He said he is “not convinced the chance of ASI doom is 100%,” while maintaining that the possibility is real and uncertain.

For him, uncertainty itself is an important reason to reconsider the pace of development. He argued that aggressively pursuing artificial superintelligence in the near term can be viewed as irresponsible when the consequences remain difficult to predict.

His argument therefore falls between two increasingly visible positions in the AI debate.

One side emphasizes potentially catastrophic risks from advanced AI and calls for slower development. Another side argues that predictions of AI-driven extinction are exaggerated and that continued technological progress should not be unnecessarily restricted.

His Work Was Connected to Faster AI Hardware

O'Callahan did not directly train frontier AI models. His work focused on software and tools related to hardware chip design.

The connection to AI development became increasingly important as the industry began demanding more powerful computing infrastructure.

More efficient AI chips can reduce the cost and time required to train and operate advanced models. That means engineering improvements at the hardware level can indirectly accelerate the development and deployment of increasingly capable AI systems.

O'Callahan eventually viewed this connection as significant enough to influence his decision to leave Google.

Google's AI Chip Ambitions Extend to Space

O'Callahan's resignation comes as Google is also exploring unconventional approaches to expanding AI computing infrastructure.

Under Project Suncatcher, Google has outlined plans to test AI chips in orbit. A prototype satellite carrying Google's Trillium Tensor Processing Units, or TPUs, is expected to fly on a SpaceX Transporter-18 rideshare mission.

The satellite has been developed with Planet, a company known for Earth-imaging technology.

The broader idea is to investigate whether AI data centers could eventually operate in space. One potential advantage is access to abundant solar energy, with orbital solar panels potentially receiving substantially more sunlight than solar infrastructure on Earth.

Google CEO Sundar Pichai has publicly highlighted the project, while O'Callahan's resignation statement did not specifically discuss Project Suncatcher.

Leaving Google Was Not an Easy Decision

O'Callahan described his position at Google as financially rewarding and intellectually enjoyable.

He said being “paid handsomely to solve fun puzzles” had been amazing, making the decision to leave particularly difficult.

Moving to another internal engineering team was also reportedly impractical because Google does not have other engineering teams in New Zealand that offered an obvious alternative.

O'Callahan is also a Christian and church elder in Auckland. He explained that, from his personal perspective, continuing to ignore the potential impact of his work would conflict with his religious principles.

A Different Position From AI Doomers

O'Callahan's resignation comes amid a broader series of departures and warnings from people connected to major AI laboratories.

Jacob Coxon recently left Anthropic and subsequently warned about the potential consequences of continuing AI development without slowing down. Former Google DeepMind researchers Rishub Jain and Bilal Chughtai have also left the company.

At the same time, prominent technology leaders have expressed differing views about the severity of AI risks.

Anthropic CEO Dario Amodei has advocated for slowing aspects of frontier AI development. Public discussion around AI safety has also involved figures including OpenAI CEO Sam Altman, Google DeepMind CEO Demis Hassabis and Tesla CEO Elon Musk.

Nvidia CEO Jensen Huang, meanwhile, has pushed back against extreme AI extinction narratives, describing such warnings as “doomsday narratives.”

O'Callahan does not appear to fit neatly into either position.

Reward Hacking and AI Misalignment Concerns

O'Callahan said some concerns previously associated with AI safety warnings have already appeared in real-world AI systems.

Among the issues he referenced were:

• Reward hacking
• AI misalignment
• Deceptive model behavior
• Rapid increases in AI capabilities
• Difficulty predicting future AI development

At the same time, he said he had not seen convincing evidence that people inside the AI industry were deliberately exaggerating AI risks to increase stock prices or prevent competitors from entering the market.

He also argued that stronger regulation is urgently needed as AI capabilities continue to advance.

O'Callahan Will Continue Working With AI

Despite leaving Google DeepMind, O'Callahan is not abandoning artificial intelligence entirely.

He plans to continue maintaining rr and Pernosco, debugging tools associated with software development. He also intends to study how AI agents can be used to debug computer code.

His stated goal is to ensure that his future work is “unambiguously pro-human.”

That suggests his departure is less about rejecting AI technology altogether and more about changing the type of AI-related work he contributes to.

Key Highlights

• Robert O'Callahan resigned from Google DeepMind over concerns about the rapid pace of AI development.
• He worked on tools for hardware chip design rather than directly developing AI models.
• He believes his work could make AI hardware cheaper and faster.
• O'Callahan considers the risk of ASI-related catastrophe real but uncertain.
• He does not believe that AI doom is necessarily inevitable.
• He has argued that stronger AI regulation is urgently needed.
• His resignation comes amid wider debate about AI safety at leading technology companies.
• He plans to continue working on debugging technology and research involving AI agents.
• His stated goal for future projects is to make them “unambiguously pro-human.”

Professional Disclaimer & Third-Party Reference

This article is based on publicly available information and statements attributed to Robert O'Callahan, Google DeepMind and other technology-industry figures. Facts and developments are referenced from standard third-party news reports, public statements, company information and other publicly available sources. Opinions regarding AI safety, artificial superintelligence and potential catastrophic outcomes are attributed to the individuals expressing them and should not be interpreted as established scientific conclusions.

Conclusion

Robert O'Callahan's departure from Google DeepMind adds another perspective to the rapidly intensifying debate over artificial intelligence safety. His position acknowledges serious AI risks without claiming that catastrophic outcomes are inevitable.

His decision also reflects a broader question facing the technology industry: how quickly should increasingly powerful AI systems be developed when their long-term consequences remain uncertain?

As AI companies continue investing heavily in models, chips and computing infrastructure, debates over safety, regulation and responsible development are likely to remain central to the future of the industry.