A new arXiv preprint, MLExpert: Towards a Generalist Agentic Model for Autonomous Machine Learning Engineering, proposes a model that goes beyond assisting humans in research and engineering. The paper describes a system that can autonomously develop and improve AI systems, an area the authors call AI for AI (AI4AI).
The paper frames this as a pathway toward recursive self-improvement, where an AI system can improve itself or other AI systems with minimal human intervention. The emphasis on "generalist" and "agentic" capabilities suggests a shift from task-specific automation to broader, autonomous engineering.
The abstract provides limited detail on the model's architecture or evaluation, and as a preprint the claims have not yet been peer-reviewed. Still, the direction reflects a growing research focus on using LLMs not just as tools but as autonomous agents in the machine learning lifecycle.