AGI Index dimension · weight 9%
Scientific discovery
Generating findings that survive independent replication.
Producing genuinely novel scientific results — hypothesis to validated finding — rather than accelerating steps within a human-directed programme. Scored on independently replicated contributions where the system did the discovering.
The argument
Evidence and counter-evidence
Both sides are published at equal weight. A framework that only records what raises a score is not measuring anything.
Raises the score
- Structural biology and materials screening produce candidates that survive experimental validation at rates that materially change research economics.
- Autonomous laboratory loops now close — propose, run, measure, revise — in narrow domains.
Holds it down
- Nearly all contributions are acceleration within a human-framed question. Framing the question remains human.
- The dimension is bounded by physical experiment, which does not speed up because the proposer got faster.
- Replication of AI-originated findings is thin, and the literature has a strong publication bias toward successes.
What the score reads from
Validated novel candidates
Independently framed research questions
Related coverage
Everything bearing on scientific discovery
Research
Context
Where this sits against the rest
- Multimodality82
- Coding81
- Tool use76
- Reasoning74
- Memory67
- Generalization66
- Planning61
- Learning59
- Autonomy53
- Social intelligence52
- Embodiment41
How these scores are producedMethodology v1.2