Shane Legg: Architecting Advanced Intelligence at Google DeepMind

Shane Legg stands as a pivotal figure in the realm of artificial intelligence, a co-founder of DeepMind Technologies and currently serving as Chief AGI Scientist at Google DeepMind. His career reflects a deep commitment to understanding and building machines capable of broad cognitive abilities, a pursuit that has shaped one of the world’s foremost AI research institutions.

Defining Artificial General Intelligence

Legg’s academic journey began with undergraduate studies in New Zealand, followed by a master’s degree focusing on Solomonoff Induction. He later pursued doctoral research at the Dalle Molle Institute for Artificial Intelligence Research, where his work on theoretical models of super intelligent machines culminated in a thesis on Machine Super Intelligence. Before his PhD and the founding of DeepMind, Legg gained experience in various software development roles, including at a big data firm and an early startup. In the early part of the new millennium, he played a significant role in re-introducing and popularizing the concept of artificial general intelligence, or AGI, alongside Ben Goertzel. This term describes an AI system that can perform virtually any cognitive task a human can, a notion that was once considered unconventional in the field. His professional focus has also extended to discussing the implications of advanced AI, including the importance of considering potential risks associated with its development.

The Genesis of DeepMind

Legg’s path intersected with Demis Hassabis in 2009 at University College London, where Legg was engaged in postdoctoral research. This meeting laid the groundwork for a significant collaboration. In 2010, Legg co-founded DeepMind Technologies with Hassabis and Mustafa Suleyman. The startup quickly garnered attention and was acquired by Google in 2014. Following its acquisition, DeepMind continued to expand its research endeavors. In 2023, the company merged with Google AI’s Google Brain division, forming Google DeepMind. In his earlier capacity as chief scientist, Legg’s responsibilities included overseeing recruitment, determining the company’s research priorities, and leading efforts in AI safety.

Advancing the Frontier of AI

Today, Legg continues his work at Google DeepMind as the Chief AGI Scientist, contributing to the company’s mission to develop advanced artificial intelligence. Google DeepMind, a British-American AI research laboratory and a subsidiary of Alphabet Inc., operates from its London headquarters and maintains research centers across multiple countries. The company has been at the forefront of numerous breakthroughs, demonstrating its commitment to pushing the boundaries of AI. Its achievements include the introduction of neural Turing machines and the development of neural network models that learn through reinforcement to excel in various video and board games. Notably, its AlphaGo program achieved a significant milestone by defeating a world champion in the game of Go. Further advancements led to AlphaZero, a more generalized program that mastered Go, chess, and shogi through self-play. Beyond game-playing, Google DeepMind has developed models for complex mathematical problems, algorithm discovery, and made substantial progress in protein folding with AlphaFold, releasing a vast database of predicted protein structures. The company is also responsible for developing prominent large language model families, such as Gemini and Gemma, alongside other generative AI models for creating images, videos, and music, all contributing to the broader pursuit of artificial general intelligence.