An Initial Look at Self-Reprogramming Artificial Intelligence

Rapid progress in deep learning research has greatly extended the capabilities of artificial intelligence technology. Conventional AI models are constrained to explicit human-designed algorithms, although a growing body of work in meta-learning, neural architecture search, and related approaches have explored algorithms that self-modify to some extent. In this paper, we develop and experimentally validate the first fully self-reprogramming AI system. Applying AI-based computer code generation to AI itself, we implement an algorithm with the ability to continuously modify and rewrite its own neural network source code.