Position Paper: Towards Transparent Machine Learning
This addresses the issue of interpretability and trust in AI systems for developers and users, though it is incremental as it builds on existing ideas of transparency.
The paper tackles the problem of opaque machine learning models by proposing transparent machine learning, where models and learning systems are represented as source code to enable direct human understanding, verification, and refinement, with the potential to enhance AI safety and security.
Transparent machine learning is introduced as an alternative form of machine learning, where both the model and the learning system are represented in source code form. The goal of this project is to enable direct human understanding of machine learning models, giving us the ability to learn, verify, and refine them as programs. If solved, this technology could represent a best-case scenario for the safety and security of AI systems going forward.