Bounded Recursive Self-Improvement
This work addresses foundational challenges for achieving artificial general intelligence, offering a blueprint for autonomous systems, though it appears incremental in its approach to self-improvement.
The paper tackles the problem of creating systems that can recursively self-improve in underspecified circumstances, achieving operational autonomy and learning complex tasks like real-time multimodal dialogue with humans through on-line observation.
We have designed a machine that becomes increasingly better at behaving in underspecified circumstances, in a goal-directed way, on the job, by modeling itself and its environment as experience accumulates. Based on principles of autocatalysis, endogeny, and reflectivity, the work provides an architectural blueprint for constructing systems with high levels of operational autonomy in underspecified circumstances, starting from a small seed. Through value-driven dynamic priority scheduling controlling the parallel execution of a vast number of reasoning threads, the system achieves recursive self-improvement after it leaves the lab, within the boundaries imposed by its designers. A prototype system has been implemented and demonstrated to learn a complex real-world task, real-time multimodal dialogue with humans, by on-line observation. Our work presents solutions to several challenges that must be solved for achieving artificial general intelligence.