How AI can help us share intuitive knowledge | Dan Shipper

Key Concepts

• Tacit Knowledge: Knowledge that is difficult to transfer to another person by means of writing it down or verbalizing it. Often based on experience and intuition.
• Explicit Knowledge: Knowledge that can be readily articulated, written down, and shared.
• Neural Networks/Language Models: AI systems capable of learning from data and performing tasks like diagnosis or providing expert advice.
• Knowledge Transfer: The process of conveying knowledge from one entity (e.g., an expert) to another (e.g., a machine or another person).
• Collaboration & Progress: The idea that societal advancement relies on the sharing and building upon existing knowledge.

---

The Limitations of Explicit Knowledge and the Rise of AI

The speaker discusses a historical trend of seeking “explicit definitions and scientific explanations” for phenomena. This pursuit is driven by the inherent advantage of being able to write things down – making knowledge easily disseminated and fostering societal progress. The core argument is that progress historically has relied on the “spreading explanations” that explicit knowledge allows. However, this approach encounters a fundamental limitation when dealing with complex domains where crucial knowledge isn’t easily codified. Specifically, the speaker highlights areas where knowledge is “parts of the world that you can’t write down explicitly,” creating a barrier to collaboration and advancement.

Capturing and Distributing Tacit Knowledge with Neural Networks

The central proposition is that neural networks offer a novel solution to this problem. They provide a mechanism to capture and distribute “intuitive experience or intuition” – what is referred to as tacit knowledge – that traditionally remains inaccessible. This tacit knowledge, built up through experience, is difficult to articulate into a set of explicit rules. The speaker emphasizes that this is particularly relevant in fields reliant on expert judgment.

The Example of Expert Clinical Diagnosis

A key example provided is expert clinical diagnosis. The speaker asserts that “the best clinicians in the world know something about how to deal with patients that they can’t write down.” This knowledge is “trapped in their head” and cannot be easily embodied in a rule-based system. Neural networks, specifically language models and broader AI systems, offer a way to externalize this intuition and embed it into a “tool” accessible to a wider audience.

Democratizing Access to Expertise

The ultimate benefit highlighted is the potential to “allow anyone in the world to access for example the best clinician in the world,” even without being able to explicitly define the clinician’s expertise. This represents a significant shift towards democratizing access to specialized knowledge and potentially improving outcomes in fields where tacit knowledge is paramount. The speaker doesn’t provide specific data or statistics, but the implication is that this technology can bridge the gap between expert knowledge and broader application.

Logical Connections & Synthesis

The argument progresses logically from identifying the limitations of relying solely on explicit knowledge, to proposing neural networks as a means of capturing tacit knowledge, and finally illustrating the potential benefits through the example of medical diagnosis. The core connection is the idea that AI can act as a vessel for transferring and distributing knowledge that was previously inaccessible due to its uncodifiable nature.

The main takeaway is that AI, particularly neural networks and language models, represents a paradigm shift in how we approach knowledge transfer and collaboration, enabling us to leverage the power of intuition and experience in a scalable and accessible way.