Adrià Garriga Alonso

Name:	Adrià Garriga Alonso
Affiliation:	University of Cambridge
E-mail:	(only provided to users who are logged into talks.cam)
Last login:	6 Jun 2022, 2:03 a.m.

Public lists managed by Adrià Garriga Alonso

• Engineering Safe AI
• Machine Learning @ CUED

Talks given by Adrià Garriga Alonso

Obviously this only lists talks that are listed through talks.cam. Furthermore, this facility only works if the speaker's e-mail was specified in a talk. Most talks have not done this.

• Neural Tangent Kernel
• Goals vs Utility Functions
• Ambitious Value Learning
• Embedded Agency
• Comprehensive AI Services
• Dynamic Safe Interruptibility for Decentralized Multi-Agent Reinforcement Learning
• Measuring and avoiding side effects using relative reachability
• Approaches to avoiding negative side effects
• Logical Induction: a computable approach to logical non-omniscience
• Last term summary + discussion of topic importance
• Counterargument to CIRL, and Safely Interruptible Agents

Talks organised by Adrià Garriga Alonso

This list is based on what was entered into the 'organiser' field in a talk. It may not mean that Adrià Garriga Alonso actually organised the talk, they may have been responsible only for entering the talk into the talks.cam system.

• Two Approximate Sampling Methods for Bayesian Deep Learning
• Can Machines Read our Minds?
• How useful is quantilization for mitigating specification-gaming?
• Misleading meta-objectives and hidden incentives for distributional shift
• Causal Reasoning from Meta-reinforcement Learning
• Inverse Game Theory
• Goals vs Utility Functions
• Who do we want to control human-level AI?
• Bayesian Theory of Mind: Modeling Joint Belief-Desire Attribution
• Ambitious Value Learning
• Machine Theory of Mind
• Embedded Agency
• Observation and Intervention Incentives in Causal Influence Diagrams: Towards an Understanding of Powerful Machine Learning Systems
• Comprehensive AI Services
• Incomplete Contracting and AI Alignment
• The Algorithmic Foundations of Differential Privacy (Chapters 1 and 2)
• Dynamic Safe Interruptibility for Decentralized Multi-Agent Reinforcement Learning
• Measuring and avoiding side effects using relative reachability
• Interpretable Machine Learning
• Scaling inverse reinforcement learning for human-compatible AI
• Motivation for this group, Goodhart's Law
• Approaches to avoiding negative side effects
• AI Safety via Debate
• AI Safety Gridworlds: Is my agent 'safe'?
• Logical Induction: a computable approach to logical non-omniscience
• Decision Boundary Geometries and Robustness of Neural Networks
• Decision Theory for AI safety
• Safe Exploration in Reinforcement Learning
• Amplification and dialogue as mechanisms for safe advanced AI
• Last term summary + discussion of topic importance
• Counterargument to CIRL, and Safely Interruptible Agents
• Reinforcement learning with a corrupted reward function
• Deep Reinforcement Learning from Human Preferences
• An introduction to adversarial attacks and defences