DeepMind philosopher reveals strain of AI's wartime commercial race

Google DeepMind has built its artificial intelligence strategy around a framework developed by an in-house philosopher, revealing how the company is attempting to manage the ethical risks of systems that now carry much of its corporate future. Iason Gabriel, who joined the lab in 2017, was for a time the only philosopher working inside a frontier AI company, tasked with anticipating the ethical fallout of technologies his colleagues were racing to build.

The commercial stakes for Google are immense. Artificial intelligence is the fastest-growing industry globally, and DeepMind's chief, Demis Hassabis, has described the competitive landscape as "wartime." As the lab trades on landmark scientific achievements like AlphaFold and Isomorphic Labs, investors and markets are watching whether ethical considerations can survive the immense pressure to deploy products quickly.

Gabriel's work attempts to bridge a historical divide in the field between "AI safety," which fears a rogue superintelligence, and "AI ethics," which focuses on current harms like biased facial recognition. In a 2020 paper, he argued that while programming a machine to follow values is difficult, choosing which values to program in a deeply divided world is harder still.

This culminated in a 267-page report on AI agents that redefined how the company views alignment. Gabriel and his team argued that alignment is not simply about a machine obeying instructions, but rather a four-way relationship involving the AI, the user, the developer, and society.

That framework has already influenced Google's products by identifying specific failure modes. An assistant trained solely to please a user might simply flatter them, a dynamic Gabriel calls "social reward hacking." Partly as a result of his work, Google's models are now specifically trained not to pretend to be people.

The durability of these guardrails remains an open question as commercial pressures mount. Helen King, who sets DeepMind's responsible-AI strategy, compared the challenge to making a knife: "a maker cannot control how it is used, but can cover the blade and warn people."

Gabriel, a "card-carrying humanist," expects artificial intelligence to be as transformative as the Industrial Revolution. However, he offers a warning rooted in history, noting that "for many who lived through it, things got worse before they got better."