Key Uncertainties & Cruxes
What Are Cruxes?
Section titled “What Are Cruxes?”A crux is a question where:
- Different answers lead to significantly different conclusions or priorities
- The question is (at least somewhat) empirically resolvable
- Reasonable people currently disagree on the answer
Identifying cruxes helps:
- Focus research on the most decision-relevant questions
- Enable productive disagreement (find what actually matters)
- Track how evidence should update views over time
Cruxes by Domain
Section titled “Cruxes by Domain”Risk Assessment Cruxes
Section titled “Risk Assessment Cruxes”Key uncertainties about which risks matter most and how severe they are:
| Domain | Page | Key Questions |
|---|---|---|
| Accident Risks | Technical failures, misalignment | Will we get warning signs? Can we verify alignment? |
| Misuse Risks | Intentional harm, weapons | How much does AI amplify malicious actors? |
| Structural Risks | Concentration, racing, lock-in | Are racing dynamics inevitable? |
| Epistemic Risks | Truth, knowledge, trust | Can verification keep pace with generation? |
Solution Cruxes
Section titled “Solution Cruxes”Key uncertainties about what interventions work:
| Domain | Page | Key Questions |
|---|---|---|
| Solutions | Epistemic & coordination tools | Can AI defense match AI offense? |
Cross-Cutting Cruxes
Section titled “Cross-Cutting Cruxes”Some questions cut across multiple domains:
Timelines
Section titled “Timelines”- When will transformative AI arrive?
- How much does this affect which interventions matter?
Alignment Difficulty
Section titled “Alignment Difficulty”- Is alignment a hard technical problem or an engineering challenge?
- Do current techniques scale?
Coordination
Section titled “Coordination”- Can labs/governments coordinate effectively?
- Do racing dynamics prevent adequate safety margins?
Warning Signs
Section titled “Warning Signs”- Will we get clear signals before catastrophic capabilities?
- Can evaluations provide meaningful safety guarantees?
How to Use This Section
Section titled “How to Use This Section”For prioritization:
- Identify which cruxes most affect your views
- Focus learning on resolving your key uncertainties
For research:
- Target empirical work at resolvable cruxes
- Design studies that could shift community views
For dialogue:
- Find where you actually disagree with others
- Avoid arguing past each other on downstream conclusions