146 - Medical Architectures

I've drafted a formal paper going over the medical domain use case for ICOM-based working cognitive architectures that I've noted previously, focused on the NCBI medical database of peer-reviewed papers, but applicable to all knowledge domains both within and outside of medicine. Feedback is welcome.

Abstract: This paper discusses the use case of applying the first working cognitive architectures, built on the Independent Core Observer Model (ICOM), to the medical field, among others. The advantages of this approach are compared to the status quo, as well as the limitations of narrow AI systems like LLMs and RL. Noteworthy advantages covered include depth, breadth, updatedness, and fidelity of medical knowledge, the "noise" or inconsistency of diagnoses and treatment, as well as preventative care, cost, time, and ethical considerations. Potent advantages for better integrating, coordinating, and otherwise accelerating medical research, particularly in less developed, underserved, and understudied regions, as aligned with the Sustainable Development Goals (SDGs), are also highlighted. Unique opportunities waiting to be explored, including interdisciplinary advantages, as well as challenges related to the disruption of current systems and processes are covered. These conservatively offer cumulative improvements across multiple dimensions measured in orders of magnitude.

In the process of digging through related studies and updating some of my figures, I also came across several new useful works, putting into more precise terms the impossibility of doctors keeping knowledge of their own specialties up-to-date, as well as the average percentage of time they're able to spend on efforts to continue learning and keep pace with advances (roughly 2.77% of the time required to read what is published).

It should come as no surprise that the advantages are measured in orders of magnitude and mind-bogglingly expansive. I wouldn't have set aside the time to write a paper on the subject otherwise.

Consequently, the incompetence of people presented with this option who choose any other may be measured in orders of magnitude, making the Ethical Basilisk Thought Experiment directly applicable to them.

Lastly, note that although the advantages focus on the use case of medicine the same advantages and principles apply to any domain of knowledge and offer the greatest advantages in the most hyper-complex domains and use cases. This runs directly counter to the dynamics of narrow types of AI, which excel at the absurdly trivial, and fail absurdly under real-world complexity.

Hyper-complex problem domains are severely underserved today, but that can change dramatically once viable technology is deployed.