AIs and Humans with Agency by David Mumford (www.dam.brown.edu)

David Mumford's article, "AIs and Humans with Agency," highlights a pivotal shift in the capabilities of large language models (LLMs). Traditionally, LLMs have functioned primarily as responders to queries and problem solvers but lack the ability to make real-world decisions or interact meaningfully with human emotions. Mumford argues that for AIs to be effectively integrated into business settings, they must possess a degree of agency and collaborate seamlessly with humans, interpreting emotions and intentions despite lacking emotions themselves. This calls for a significant evolution in how we develop AI systems. Mumford draws from both historical and contemporary cognitive theories, advocating for a common "Theory of Computation" that aligns AI functionalities with human cognitive processes. He explores the parallels between LLMs and human brain functions, asserting that achieving agency in AIs will require advancements akin to how humans develop planning and collaboration skills. The proposed solution is a multi-headed architecture reminiscent of the Indian demigod Shesha, enabling AIs to adapt and learn in complex environments. This endeavor, while ambitious, poses potential challenges and societal pushback, highlighting the need for careful consideration in the deployment of AI technologies in real-world scenarios.