Research lab or companyUser-requested source
Scale Labs
research lab/company
Executable agent benchmarks, human-feedback research, data, safety, and evaluation.
AI Thought Leadership · Individual
Demis Hassabis
Neuroscience, reinforcement learning, differentiable memory, and search-based decision systems.
Type
Individual (individual research leader)
Directory section
User-requested source
What to monitor
Technical publications from teams he leads, talks that make testable research predictions.
Directory last reviewed
August 16, 2026
Why this source is monitored
Demis Hassabis's authored research connects the constructive role of hippocampal memory to deep reinforcement learning, learned external memory, neuroscience-informed AI, and self-play with tree search. Credit attaches only to papers on which he is an author. Leadership of DeepMind does not imply authorship of every laboratory result.
Official links
Reviewed topic lineages
Caveat and non-endorsement
Strategic perspective is not a substitute for the lab's technical evidence.
Evidence boundary. Human neuroscience supplies a conceptual precursor. Game environments provide controlled decision experiments. Neither alone establishes open-world general intelligence.
Inclusion in this directory is a monitoring decision by Implement Agentic. It is not an endorsement, a partnership, a client relationship, or a claim that this person or organization endorses us. Product claims and personal forecasts must be traced to technical evidence before they are used as fact.
How we weight this kind of source
- Primary research: papers, technical reports, system cards, benchmarks, official repositories, university research, or government evaluations.
- Open ecosystem: code, weights, datasets, model cards, evaluation tooling, and reproducibility work. Verify license and provenance separately.
- Expert synthesis: interpretation by a named researcher or practitioner. Link through to the underlying evidence.
- Market signal: startup, investor, or portfolio activity. Use to detect direction of interest, not as proof of capability.
- Publication/index: discovery infrastructure. A listing establishes publication or preprint status, not correctness.
Other sources in this section
Research lab or companyUser-requested source
Anthropic Research
research lab/company
Frontier-model safety, interpretability, agents, human-use studies, and system-level trust.
Research lab or companyUser-requested source
OpenAI Research
research lab/company
Frontier models, agents, multimodal systems, safety, and AI-for-science programs.