Shane Legg

Shane Legg CBE is a New Zealand computer scientist and machine-learning researcher, born 1973. He is a co-founder of Google DeepMind and serves as the lab's Chief AGI Scientist, the role responsible for long-horizon research direction toward artificial general intelligence. As of May 2026, he is one of three original DeepMind co-founders alongside Demis Hassabis and Mustafa Suleyman, the only one of the three to have remained at the lab continuously since the September 2010 founding, and one of the most-cited voices on AGI timeline forecasts in industry coverage.

At a glance

• Education: BCMS, University of Waikato (1996); MSc in computer science, University of Auckland (1996), with a thesis on Solomonoff induction supervised by Cristian S. Calude; PhD in computer science, Dalle Molle Institute for Artificial Intelligence (IDSIA), University of Lugano (2008), supervised by Marcus Hutter, thesis Machine Super Intelligence, defended June 17, 2008.
• Current role: Co-founder and Chief AGI Scientist, Google DeepMind, since 2010 (title since the April 2023 Brain merger; previously Chief Scientist).
• Key contributions: Co-founder of DeepMind (September 2010); co-author with Marcus Hutter of Universal Intelligence: A Definition of Machine Intelligence (2007); the Machine Super Intelligence dissertation (2008); the popularization of the term "Artificial General Intelligence" with Ben Goertzel in the early 2000s.
• Awards: Commander of the Order of the British Empire in the 2019 Birthday Honours, for services to the science and technology sector and to investment; Singularity Institute Prize (2008); TIME 100 Most Influential People in AI (2023).
• X / Twitter: @ShaneLegg
• Personal site / blog: vetta.org
• Wikipedia: Shane Legg
• Google DeepMind bio: deepmind.google/about

Origins

Legg was born in 1973 in New Zealand and attended Rotorua Lakes High School. He studied at the University of Waikato in Hamilton, taking a Bachelor of Computing and Mathematical Sciences in 1996, and at the University of Auckland for a Master of Science in computer science the same year. The Auckland thesis on Solomonoff induction, supervised by Cristian S. Calude, set the technical direction for the rest of his career: the mathematical formalization of intelligence, induction, and learning under uncertainty.

After the master's degree he spent several years in software development, working at the big-data firm Adaptive Intelligence and at WebMind, the early AGI startup founded by Ben Goertzel in the late 1990s. The WebMind period is the documented start of the Goertzel collaboration that would later produce, in the early 2000s, the popularization of the term "artificial general intelligence" as the title for the 2007 edited volume that has since given the field its modern name.

Legg returned to academia in the mid-2000s and enrolled for doctoral work at the Dalle Molle Institute for Artificial Intelligence (IDSIA), affiliated with the University of Lugano in Switzerland, under the supervision of Marcus Hutter, with Jürgen Schmidhuber as co-advisor. The thesis, Machine Super Intelligence, was defended on June 17, 2008. It is one of the earliest scholarly book-length treatments of superintelligence, surveying definitions of intelligence, formalizing a universal measure of intelligence, and working through the AIXI model. The Singularity Institute awarded the dissertation its $10,000 prize that year.

Career

After the doctorate Legg held a postdoctoral fellowship in finance at the Università della Svizzera italiana (USI) in Lugano, before moving to London in 2009 for a research-fellowship position at the Gatsby Computational Neuroscience Unit at University College London. He met Demis Hassabis at the Gatsby Unit the same year, and the two began the conversations that would lead to the founding of DeepMind.

Legg co-founded DeepMind Technologies in London in September 2010 with Hassabis and Mustafa Suleyman. The premise was that progress toward AGI would come from combining deep learning, reinforcement learning, and ideas from neuroscience. Early work focused on Atari 2600 video-game agents and the deep Q-network research that produced "Human-level control through deep reinforcement learning" in Nature in February 2015. Legg held the title of Chief Scientist and was the lab's principal in-house authority on AGI definitions and long-horizon research direction.

Google acquired DeepMind in January 2014, with the lab continuing as a semi-autonomous subsidiary inside Google and, from 2015, the broader Alphabet holding company. Legg remained in his research-leadership role across the acquisition and the lab's expansion through the AlphaGo (2015 to 2016), AlphaZero (2017), and AlphaFold (2020 and 2024) periods. He was appointed Commander of the Order of the British Empire in the 2019 Birthday Honours, for services to the science and technology sector and to investment, the same honors year in which Suleyman departed DeepMind to found Inflection AI.

In April 2023, Alphabet merged DeepMind with the Google Brain research organization to form Google DeepMind. Legg's role was rebranded from Chief Scientist to Chief AGI Scientist. He has remained active in DeepMind's interactive-environment research, including the SIMA program announced in March 2024, which trains Gemini-based agents in 3D video-game environments as a stepping stone to embodied AGI. In October 2023 on the Dwarkesh Patel podcast he stated a 50 percent probability of AGI by 2028, consistent with the log-normal distribution he first published on the vetta.org blog in 2011 with a mode of 2025.

Affiliations

• Adaptive Intelligence; WebMind: Software developer, late 1990s; the WebMind period was under Ben Goertzel.
• IDSIA / University of Lugano: PhD candidate, completed 2008 (advised by Marcus Hutter).
• Università della Svizzera italiana (USI): Postdoctoral fellow in finance, 2008.
• University College London: Research fellow, Gatsby Computational Neuroscience Unit, 2009.
• Google DeepMind: Co-founder and Chief Scientist from September 2010, currently Chief AGI Scientist, to present.

Notable contributions

Legg's published record concentrates on the formal definition of intelligence, the popularization of the AGI research agenda, and his co-founder role at DeepMind.

• Universal Intelligence: A Definition of Machine Intelligence (2007). The Legg-Hutter paper in Minds and Machines formalizes intelligence as an agent's ability to achieve goals across a wide range of environments, weighted by Kolmogorov complexity. It is widely cited as the canonical mathematical statement of universal intelligence.
• Machine Super Intelligence (2008). The IDSIA doctoral dissertation, supervised by Marcus Hutter and co-advised by Jürgen Schmidhuber. Among the earliest book-length scholarly treatments of superintelligence; awarded the Singularity Institute's $10,000 prize that year.
• Popularization of the term "Artificial General Intelligence" (early 2000s). Coined the title for the 2007 Ben Goertzel and Cassio Pennachin edited volume Artificial General Intelligence, launching the term as the standard label for the research agenda. The phrase had been used in passing by Mark Gubrud in 1997, but its present-day adoption traces to the Goertzel and Legg framing.
• Co-founding of DeepMind Technologies (September 2010). With Demis Hassabis and Mustafa Suleyman in London. Subsequently chief scientist and chief AGI scientist for fifteen years through the AlphaGo, AlphaZero, AlphaFold, and Gemini periods.
• AGI timeline forecasts. Long-running public commentary that has converged on a 50 percent probability of AGI by 2028, with a published log-normal distribution from a 2011 blog post placing the mode at 2025. Restated on the Dwarkesh Patel podcast in October 2023.
• SIMA (March 2024). The DeepMind generalist agent for 3D virtual environments, trained on Gemini, framed publicly as a stepping stone to embodied AGI.
• 2023 Statement on AI Risk. Among the senior signatories of the Center for AI Safety's single-sentence statement on extinction risk from AI.
• Awards. Commander of the Order of the British Empire (2019); Singularity Institute Prize (2008); TIME 100 Most Influential People in AI (2023).

Investments and boards

The CBE citation in the 2019 Birthday Honours references "services to the science and technology sector and to investment," which industry coverage has read as recognition of Legg's investor activity in the British and European AI ecosystem alongside the DeepMind operating role. The named transactions are not publicly itemized, and no public personal angel-investor activity in AI, semiconductors, datacenters, software, or energy is on record at a level suitable for inclusion in this section as of May 2026.

• Google DeepMind (AI): Co-founder and Chief AGI Scientist, 2010 onwards. Subsidiary of Google from January 2014; merged with Google Brain in April 2023.

Network

Legg's longest-running professional relationships are inside DeepMind. Demis Hassabis, the chief executive and co-founder he met at the Gatsby Computational Neuroscience Unit in 2009, is the principal collaborator across the founding period and the fifteen years that followed. Mustafa Suleyman, the third co-founder, departed in 2019 to found Inflection AI and now leads Microsoft AI. John Jumper, the AlphaFold research lead and 2024 Nobel Laureate, has been a senior research-bench colleague since 2017. David Silver, the former VP of reinforcement-learning research, was a peer at DeepMind from 2010 until his January 2026 departure to found Ineffable Intelligence. Tim Rocktäschel, the former director of the Open-Endedness group, departed in early 2026 to co-found Recursive Superintelligence.

Outside DeepMind, Legg's strongest academic relationship is with Marcus Hutter, his IDSIA doctoral advisor and co-author on the universal-intelligence paper. Hutter holds positions at the Australian National University and as a senior researcher at Google DeepMind, sustaining the partnership across more than two decades. Jürgen Schmidhuber, his IDSIA co-advisor, is one of the founding figures of modern deep learning. Ben Goertzel, his WebMind colleague, leads SingularityNET.

Position in the field

As of May 2026, Legg is the only one of the three original DeepMind co-founders to have remained at the lab continuously through the Google acquisition (January 2014), the AlphaFold breakthroughs, the Brain merger (April 2023), and the Gemini cycles. The continuous tenure puts him alongside Hassabis as the longest-tenured senior leader of any frontier AI laboratory.

His scientific footprint is concentrated upstream of the production-research line that has produced DeepMind's headline results. The Legg-Hutter universal-intelligence definition (2007) and the Machine Super Intelligence dissertation (2008) are widely cited as foundational to the modern AGI research agenda, and the popularization of the term "Artificial General Intelligence" gave the field its current name. Industry coverage positions him as the in-house authority on AGI definitions and timelines, distinct from Hassabis's external-facing executive role and from production-research leaders like John Jumper on AlphaFold or the now-departed David Silver on reinforcement learning.

The 2019 CBE, the TIME 100 Most Influential People in AI (2023) listing, and the recurring inclusion in major AGI-timeline interviews place Legg among the small group of senior AI researchers whose public statements on AGI probability function as standard reference points.

Outlook

Open questions over the next 6 to 18 months:

• Successor research direction. Whether the SIMA generalist-agent line and DeepMind's broader interactive-environment research produces public capability milestones consistent with Legg's stated AGI-by-2028 forecast.
• AGI timeline restatement. Whether the 50-percent-by-2028 framework holds, tightens, or loosens given the trajectory of Gemini 3, GPT-5.5, and Claude Opus 4.7-class systems through 2026 and 2027.
• Long-term role at Google DeepMind. Whether Legg remains at the lab through the next research cycle, given the 2025 to 2026 pattern of senior researchers leaving frontier labs and the Silver and Rocktäschel departures from DeepMind specifically.
• AGI-definition publishing. Whether Legg or DeepMind publishes a refined formal definition or measurable benchmark suite for AGI, given that the universal-intelligence framework has remained the primary scholarly reference for nearly two decades.
• Coordination with Marcus Hutter. Whether the long-running Legg-Hutter research partnership produces additional jointly-authored work as Hutter continues at Google DeepMind alongside the Australian National University position.

Sources

• Shane Legg. Wikipedia biographical entry; covers education, the IDSIA dissertation, the founding of DeepMind, the CBE, and AGI commentary.
• Universal Intelligence: A Definition of Machine Intelligence. The 2007 Legg-Hutter paper in Minds and Machines, the canonical formal definition of universal intelligence.
• Machine Super Intelligence. The 2008 IDSIA doctoral dissertation hosted on Legg's personal site.
• Shane Legg (DeepMind Founder): 2028 AGI, superhuman alignment, new architectures. The October 26, 2023 Dwarkesh Patel podcast interview on AGI timelines, alignment, and architectures.
• Shane Legg: The 100 Most Influential People in AI 2023. TIME magazine entry on the AGI research and DeepMind role.
• SIMA: Generalist AI agent for 3D virtual environments. The March 2024 DeepMind announcement of the SIMA agent.
• Statement on AI Risk. The Center for AI Safety's 2023 single-sentence statement on AI extinction risk.
• Artificial General Intelligence. The 2007 Goertzel and Pennachin volume that established the AGI label.
• Marcus Hutter. Wikipedia entry for Legg's IDSIA doctoral advisor and long-running co-author.
• vetta.org. Legg's personal site, hosting the dissertation, the universal-intelligence paper, and the 2011 AGI-timeline post.
• Photo: Shane Legg, 2023, CC BY-SA 4.0 Libor.burian.