Current Projects
01
AI Trust Management
The focus of this work is creating infrastructure for trust propagation and audit trails, ensuring AI decision support meets business and compliance requirements.
Read the whitepaper here: The AI Trust Chain
Follow the project on GitHub: https://github.com/mossrake/ai-trust-chain
02
Structured Reasoning for Complex Investigations
This project aims to create a comprehensive framework for investigative support that tracks multiple contradictory hypotheses over noisy/asynchronous evidence. A key design point is minimizing token consumption for long-running reasoning in complex investigations.
Read the whitepapers here: Living in the Contradiction
Structured Reasoning for Complex Investigations
Follow the project on GitHub: https://github.com/mossrake/async-reasoning-engine
03
AI-enabled Smart Home
The internet made the smart home possible. AI is the next step. In collaboration with an OEM incubator, this work explores anchor tenancy in the smart home ecosystem by safely AI-enabling systems and products we interact with on a daily basis.
04
Beyond Constitutional AI
This work explores the application of narrative psychology and YN Harari's notion of AI as "alien intelligence" to shape Language Model behavior for AI coexistence.
Read the whitepaper here: Beyond Constitutional AI
Follow the project on GitHub: https://github.com/mossrake/beyond-constitutional-ai
05
Episodic Compression to Latent Variables for Sample-Efficient Intelligence
This research explores whether rather than from embodied experiences we can construct a sample-efficient continuous learning system using narratives with episodic compression to latent variables and a 2-layer learning architecture.
Read the whitepapers here: Continuous Learning From Narratives
Robot Learning Experiment using Episodic Compression
Follow the project on GitHub: https://github.com/mossrake/learning-system