Shane Legg
The theoretical pioneer who first defined AGI as a measurable goal
Shane Legg is a co-founder of DeepMind and one of the earliest researchers to transform artificial general intelligence (AGI) from a philosophical concept into a formalizable, measurable scientific goal. He completed his doctoral research at Victoria University of Wellington in New Zealand, proposing a mathematical definition of universal intelligence based on Kolmogorov complexity — universal intelligence is an agent's weighted average performance across all environments. This definition provided a theoretical foundation for AGI research. In 2010 he co-founded DeepMind with Demis Hassabis and Mustafa Suleyman, combining neural networks with reinforcement learning to pioneer the modern AGI research institution. He was also one of the earliest researchers to systematically discuss AGI risk, warning about the existential threat that superintelligence might pose to humanity long before the GPT era.
Methodologies
- Vague Goal Formalization Method - Transform intuitive, philosophical goals into mathematically operational definitions to lay a debatable foundation for long-term research
- Preemptive Risk Analysis Method - Systematically analyze the potential risks of technology before it matures, securing a time window for safety measures research in advance
Key decisions and timeline
- Published Formal Definition of Universal Intelligence Paper - Formalizing vague problems is an important first step in scientific progress
- PhD Thesis: Machine Super Intelligence - Academic seriousness can lend legitimacy to niche areas of concern
- Co-Founded DeepMind - Mission-driven research institutions can achieve technical breakthroughs that commercially-oriented companies cannot match
Beliefs and mental models
- Belief 1 - Legg and Marcus Hutter proposed a formal definition of universal intelligence: an agent's universal intelligence is its average performance across all computable environments, weighted by the inverse of each environment's complexity (Kolmogorov complexity). This transformed AGI from a philosophical concept into a mathematically tractable object.
- Belief 2 - Legg warned as early as 2008 in his doctoral thesis and multiple interviews that if humans build AI systems surpassing human intelligence without ensuring alignment with human values, the consequences could be catastrophic. He was one of the early advocates for AI safety research as a serious academic field.
- Belief 3 - Legg believes the human brain is the only known instance of general intelligence to date. Understanding how the brain learns, processes information, and achieves cross-domain generalization is the most important source of inspiration for building AGI. DeepMind's founding mission — building general AI using neuroscience principles — embodies this belief.
- Model 1
- Model 2
- Model 3