Historical overview
The concept was formalised by I. J. Good in 1965 as the 'intelligence explosion' and developed in depth by Nick Bostrom's 2014 book 'Superintelligence: Paths, Dangers, Strategies'.
Scientific basis
Three commonly distinguished paths: speed superintelligence (a human-level mind running much faster), collective superintelligence (many human-level minds coordinated), and quality superintelligence (qualitatively better cognition).
Strengths
- Could in principle solve problems intractable to human science
- Self-improvement loops may compound capability rapidly
Limitations
- Entirely hypothetical
- Existential risk if value alignment fails
Relationship to other intelligence systems
AGI
Most credible pathways to superintelligence pass through AGI.
Collective Intelligence
Collective superintelligence is a non-individual variant.
Future implications
Whether superintelligence is achievable, desirable, or controllable is the central question of long-term AI safety research.