In 2017, a political philosopher named Iason Gabriel received an unconventional suggestion from a friend: to apply for a position at DeepMind, the renowned AI research division of Google based in London. This recommendation seemed out of place, as Gabriel was a dedicated academic, interested in political theory, Vipassana meditation, and passionate about rock climbing. The son of a Greek management professor and a British documentary filmmaker, he balanced his time between teaching at Oxford, where he held a fellowship at St John’s College, and humanitarian efforts for the United Nations in regions like Sudan and Lebanon.
DeepMind, which had been purchased by Google for $650 million in 2014, was recognized as the forefront of AI research. Its fame was largely attributed to breakthroughs like AlphaGo, which triumphed over human Go champion Lee Sedol in a groundbreaking match in 2016, showcasing the sophisticated capabilities of AI in understanding and mastering extremely complex games.
While Gabriel was aware of the stir created by AlphaGo, he couldn't fathom why a company focused on AI gamification would need someone with his ethical expertise. However, he soon realized that DeepMind aimed to achieve far more than simply developing game-playing AIs. Founded by Demis Hassabis, Shane Legg, and Mustafa Suleyman in 2010, the company had ambitions to create artificial general intelligence (AGI)—intelligent systems with cognitive capabilities equal to or surpassing human intelligence. This notion was once deemed overly ambitious, but the founders remained steadfast in their goal: to "solve intelligence and then everything else."
Recognizing the significant implications of developing AGI, Legg had long maintained that society must begin contemplating the consequences of such advancements well in advance. He believed that involving ethicists in the process was crucial, especially as creating AGI might impact humanity at large.
Upon joining DeepMind as its only working philosopher at the time, Gabriel's academic background provided a unique perspective amid a primarily engineering-driven field. Over the following years, he produced influential work that identified and often anticipated ethical dilemmas linked to the rise of large language models (LLMs) and their societal impact.
As AI expert Dylan Hadfield-Menell noted, Gabriel arrived at a pivotal moment, successfully broadening the ethical discourse without undermining previous technological endeavors. He has been vocal about the need for not only new technical languages but also novel ways of thinking about our technology interactions and the ethical frameworks surrounding them.
Historically, there existed two distinct philosophies when discussing AI's social implications: AI safety and AI ethics. The AI safety faction mirrored DeepMind's founders in believing that human-level machine intelligence was not only attainable but imminent. Their focus was to ensure that these systems were aligned with human values, addressing what has become known as the alignment problem—a central concern in developing autonomous systems.
In contrast, proponents of AI ethics highlighted more immediate societal risks, emphasizing the need for social solutions to address issues such as algorithmic bias. A prominent example was researcher Joy Buolamwini's Gender Shades project, which unveiled biases in facial-recognition software. Both factions sometimes exhibited mutual disdain, with each faction arguing about whether the focus should hinge on future existential risks or present-day impacts.
Gabriel's first major research project at DeepMind in 2020 tackled the alignment challenge while exploring its ethical and political ramifications. He argued that not only was aligning AI with a set of values difficult, but selecting which values to incorporate was arguably more complicated in a world of diverse beliefs. This assertion resonated with fellow researchers, recognizing the essential need to navigate the complex dynamics of societal values.
Throughout his tenure at DeepMind, Gabriel took a proactive role in predicting potential problems associated with LLMs, publishing his analysis on their prospects well before their broader adoption. As these AI models evolved, they presented both extraordinary capabilities and significant risks that required careful consideration.
While initial skepticism about LLMs existed at DeepMind, external phenomena like the astounding success of ChatGPT in 2022 prompted a re-evaluation of the organization's stance. This chatbot's rapid rise in popularity forced Google to integrate LLM efforts with DeepMind, signaling an urgent shift in the competitive landscape.
DeepMind has historically aligned itself more with traditional research paradigms rather than startup culture, prioritizing innovation without the relentless commercial pressure that characterizes the tech industry. Yet, as AI technology becomes a focal point of global competition—including the escalating arms race in AI between nations—this stability has been challenged. With significant financial investments flooding into the AI sector, the company is now under immense pressure to deliver results.
In this evolving context, Gabriel has sharpened his focus on the extensive societal influences of AGI, anticipating that its arrival could reshape economics, politics, and even interpersonal relations. Despite expressing cautious optimism about the potential benefits of AI, he acknowledges the historical lessons of technological revolutions that often begin with disruption before yielding improvements.
Ultimately, Gabriel sees the integration of AI into society as a complex transition that demands careful navigation to harness its benefits while addressing inherent risks. As humanity stands on the brink of what could be a transformative era, grappling with AGI may compel us to confront fundamental philosophical questions about our identity and relationship with technology.
