Part 7/10:
- Alternatively, if the agent performs nearly optimally, it must possess an approximation of these causal models.
Currently, the state of causal modeling technology has not evolved to the necessary sophistication required for robust agent-based systems. DJ highlights that even sophisticated causal inference techniques remain limited in scope and often operate with painfully slow methodologies—hampering automation and scalability.
Bridging the Gap in Causal Understanding
In light of these barriers, DJ explores the methods used in contemporary causal modeling—primarily the practice of experimentation. Running controlled experiments is ideal to ascertain the outcomes of various actions, yet the volume of decision-making far outstrips the capacity to conduct these experiments.