Research Studies

Empirical investigations and theoretical studies exploring AI safety, superintelligence dynamics, and the capabilities of advanced AI systems.

Active Studies

In Progress
2025

Capability Scaling Laws in Large Language Models

Investigating the relationship between model scale, training compute, and emergent capabilities in large language models. Analyzing whether capability emergence follows predictable patterns or exhibits discontinuous jumps that could indicate threshold effects relevant to superintelligence theory.

Empirical AnalysisScaling LawsLLMs
Status: Data collection phase • Expected completion: Q3 2025
Active Research
2025

Multi-Agent Coordination in AI Systems

Examining how multiple AI agents coordinate and compete in shared environments. Investigating whether collective intelligence properties emerge that differ from individual agent capabilities, and analyzing implications for distributed superintelligence scenarios.

Multi-Agent SystemsCoordinationEmergence
Status: Analysis phase • Expected completion: Q4 2025
Literature Review
2025

Historical Analysis of AI Capability Predictions

Comprehensive analysis of AI capability predictions from 1950-2024, examining prediction accuracy, methodology quality, and common failure modes. Extracting lessons for improving current forecasting methods for advanced AI timelines and capabilities.

ForecastingHistory of AIMeta-Analysis
Status: Literature review • Expected completion: Q2 2025

Completed Studies

Completed
2024

Reasoning Capabilities in Current LLMs

Systematic evaluation of reasoning capabilities across major large language models (GPT-4, Claude, Gemini). Assessed mathematical reasoning, logical deduction, causal inference, and analogical thinking to establish baseline capabilities and identify failure modes.

CompletedBenchmarkingReasoning
View related publications →
Completed
2024

Intelligence Explosion Theory: 60-Year Synthesis

Comprehensive synthesis of intelligence explosion theory from I.J. Good (1965) through contemporary models. Analyzed evolution of concepts, identified theoretical gaps, and examined relationship between classical theory and modern AI developments.

CompletedTheorySynthesis
Read full paper →

Research Collaboration Opportunities

Interested in collaborating on any of these studies or proposing new research directions? Francis Clase welcomes partnerships with researchers, institutions, and organizations.

Discuss Research Collaboration