Polanyi’s Revenge and AI’s New Romance with Tacit Knowledge

Lately though, Polanyi’s paradox is turning into Polanyi’s revenge both in research and practice of AI. Recent advances have made AI synonymous with learning from massive amounts of data, even in tasks for which we do have explicit theories and hard-won causal knowledge.a This resistance to accept any kind of explicit knowledge into AI systems—even those operating in tasks and environments of our design—is perplexing. The only “kosher” ways of taking explicit knowledge in deep learning systems seem to be to smuggle them in through architectural biases, or feeding them manufactured examples. Anecdotal evidence hints that industry practitioners readily convert doctrine and standard operating procedures into “data” only to have the knowledge be “learned back” from that data. Even researchers are not immune—a recent paper in Nature Machine Intelligence focused on how to solve Rubik’s Cube by learning from billions of examples, rather than accept the simple rules governing the puzzle. There are policy implications too. Many governmental proposals for AI research infrastructure rely exclusively on creating (and curating) massive datasets for various tasks. The current zeal to spurn hard-won explicit (and often causal) knowledge, only to try to (re)learn it from examples and traces as tacit knowledge, is quixotic at best. Imagine joining a company, and refusing to take advice on their standard operating procedures, and insisting instead on learning it from observation and action. Even if such an approach might unearth hidden patterns in how the company actually works, it will still be a wildly inefficient way to be an employee. Similar concerns will hold for AI assistants in decision support scenarios.

Indeed, AI’s romance with tacit knowledge has obvious adverse implications to safety, correctness, and bias of our systems. We may have evolved with tacit knowledge, but our civilization has been all about explicit knowledge and codification—however approximate or aspirational. Many of the pressing problems being faced in the deployment of AI technology, including the interpretability concerns, the dataset bias concerns as well as the robustness concerns can be traced rather directly back to the singular focus on learning tacit knowledge from data, unsullied by any explicit knowledge taken from the humans. When our systems learn their own representations from raw data, there is little reason to believe that their reasoning will be interpretable to us in any meaningful way. AI systems that refuse to be “advised” explicitly are taking the “all rules have exceptions ” dictum to the “what are rules?” extreme, which flies in the face of civilizational progress, and seriously hinders explainability and contestability of machine decisions to humans in the loop.