AI Alignment

There are narrow conceptions of alignment, where you’re trying to get it to do something like cure Alzheimer’s disease without destroying the rest of the world. And there’s much more ambitious notions of alignment, where you’re trying to get it to do the right thing and achieve a happy intergalactic civilization.

But both the narrow and the ambitious alignment have in common that you’re trying to have the AI do that thing rather than making a lot of paperclips.

AI Sub-TopicsBasic Alignment Theory

AIXI
Corrigibility
Decision Theory
Embedded Agency
Fixed Point Theorems
Goodhart's Law
Inner Alignment
Instrumental Convergence
Logical Induction
Mesa-Optimization
Myopia
Newcomb's Problem
Optimization
Orthogonality Thesis
Outer Alignment
Solomonoff Induction
Utility Functions

Engineering Alignment

AI Boxing (Containment)
AI Safety via Debate
Factored Cognition
Mesa-OptimizationHumans Consulting HCH
Orthogonality ThesisImpact Measures
Iterated Amplification
Value Learning

Strategy

Utility FunctionsAI Progress
Instrumental ConvergenceAI Risk
GPTAI Services (CAIS)
AI Takeoff
AI Timelines

Other

Centre for Human-Compatible AI
Future of Humanity Institute
GPT
Machine Intelligence Research Institute
OpenAI
Ought
Research Agendas 

See also General Intelligence.

Artificial Intelligence is the study of creating intelligence in algorithms. On LessWrong, the primary focus of AI discussion is to ensure that as humanity builds increasingly powerful AI systems, the outcome will be good. The central concern is that a powerful enough AI, if not designed and implemented with sufficient understanding, would optimize something unintended by its creators and pose an existential threat to the future of humanity. This is known as the AI alignment problem.

Artificial Intelligence is the study of intelligence in algorithms. On LessWrong, the primary focus of thisAI discussion is to ensure that as humanity builds increasingly powerful AI systems, the outcome will be good. The central concern is that a powerful enough AI, if not designed and implemented with sufficient understanding, would optimize something unintended by its creators and pose an existential threat to the future of humanity. This is known as the AI alignment problem.