Artificial Intelligence is the study of creating intelligence in algorithms. AI Alignment is the task of ensuring [powerful] AI system are aligned with human values and interests. 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.

Common terms in this space are superintelligence, AI Alignment, AI Safety, Friendly AI, Transformative AI, human-level-intelligence, AI Governance, and Beneficial AI. This entry and the associated tag roughly encompass all of these topics: anything part of the broad cluster of understanding AI and its future impacts on our civilization deserves this tag.

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.

See also General Intelligence.

Basic Alignment Theory

AIXI
Coherent Extrapolated Volition
Complexity of Value
Corrigibility
Deceptive Alignment
Decision Theory
Embedded Agency
Goodhart's Law
Goal-Directedness
Gradient Hacking
Infra-Bayesianism
Inner Alignment
Instrumental Convergence
Intelligence Explosion
Logical Induction
Logical Uncertainty
Mesa-Optimization
Multipolar Scenarios
Myopia
Newcomb's Problem
Optimization
Orthogonality Thesis
Outer Alignment
Paperclip Maximizer
Power Seeking (AI)
Recursive Self-Improvement
Simulator Theory
Sharp Left Turn
Solomonoff Induction
Superintelligence
Symbol Grounding
Transformative AI
Utility Functions
Whole Brain Emulation

Engineering Alignment

Agent Foundations
AI-assisted Alignment
AI Boxing (Containment)
Debate (AI safety technique)
Eliciting Latent Knowledge
Factored Cognition
Humans Consulting HCH
Impact Measures
Inverse Reinforcement Learning
Iterated Amplification
Mild Optimization
Oracle AI
Reward Functions
RLHF
Shard Theory
Tool AI
Interpretability (ML & AI)
Value Learning

Organizations

AI Safety Camp
Alignment Research Center
Anthropic
Apart Research
AXRP
CHAI (UC Berkeley)
Conjecture (org)
DeepMind
FHI (Oxf

...

5

Inverse Rubric Optimization: A testbed for agent science

5

5

5