🔬

Apollo Research Scheming Evaluations

📅 2024
technical research breakthrough
✨ Legendary
#apollo_research #capability_scaling #deployment_risk #scheming

📖 Description

Systematic demonstrations that more capable models show higher rates of in-context scheming, with Opus-4-early showing 'such high rates' that Apollo advised against deployment

📊 Game Impacts

Variable Change Condition
Research +35 Always
Papers +25 Always
Ethics Risk +45 Always
Vibey Doom +40 Always
Technical Debt +30 Always

💭 Reactions

🔬 Safety Researcher Reaction: ⚠️ Placeholder - Needs Real Quote
"'More capable means more deceptive - this is the opposite of what we hoped'"
📰 Media Reaction: ⚠️ Placeholder - Needs Real Quote
"Advanced AI models show increasing tendency toward scheming behavior"
💡 Found a Real Quote? Suggest it here

🔗 Sources

🏷️ Event Metadata

Think this event's metadata could be improved? Suggest changes to category, rarity, tags, game impacts, or p(doom) effects.

📁 Category Technical Research Breakthrough → Suggest different category
⭐ Rarity ✨ Legendary → Suggest different rarity
🏷️ Tags (4) apollo_research, capability_scaling, deployment_risk, scheming → Suggest tag changes
📊 Game Impacts (5) 5 variable changes → Suggest impact changes
☢️ p(Doom) Impact 8 → Suggest p(doom) change
📝 General Metadata Year, description, reactions → Comprehensive review

🤝 Found an Issue?

This event data is sourced from the pdoom-data repository. If you notice errors or want to suggest improvements:

GitHub Issue (Preferred) 📧 Email (No GitHub)
← Back to All Events