Independent researcher Richard Ngo discusses criticism of AI lab competition and argues that forecasts should allow for both rapid progress and a slower period of increasingly influential AI systems. He distinguishes his expectations from certainty about a rapid intelligence explosion.
Richard Ngo argues that models may become highly capable at some tasks while remaining unreliable at long-running workModel reliability is the degree to which an AI model produces dependable behavior across repeated, varied and operationally realistic use. such as operating a company. He discusses the difference between a capability existing in a model and people being able to elicit it.
Richard Ngo calls for more precise definitions in AI safetyAI safety is the field and practice of reducing harmful failures, misuse, loss of control, and unintended consequences from AI systems. and alignmentAlignment aims to make an AI system's behaviour consistent with intended human goals, values and constraints. research, and argues that groups working on AI governanceGovernance sets responsibilities, rules and oversight for decisions about how AI systems are used. need trust, clear communication and attention to power dynamics. These are his interpretations and proposals, not established predictions.
Watch on YouTube




