Kirk Klasson

Innovation and Analogy: One of These Things Is a Lot Like Another

A while back we examined the notion that innovation, like social phenomena (see Social Subsidization and Diminishing Returns – March 2015) and venture capital (see Greed Is Good – August 2017), tends to subscribe to specific power laws. In the instance pertaining to innovation there was a strong suggestion that the richest opportunities for new discoveries would lie not within a specific domains of knowledge but at the frontiers of different knowledge domains (see Innovation: Power Laws, the Adjacent Possible and the Accumulation of Continuity – March 2017). In fact, this is not a new notion but one that has been floating around in the Knowledge Management space for the last several decades.

Last week at KDD2017 (Knowledge Discovery and Data Mining conference) a team of researchers from Carnegie Mellon and Hebrew University took the wraps off a paper, “Accelerating Innovation Through Analog Mining”, that proposes that AI can be trained on public and private data sources to discover and, by analogy, tease out the efficacy of existing solutions to unsolved problems. An analogy is a kind of verbal calculus that juxtaposes two or more things or concepts or phenomena and proposes an attribute or attributes that they have in common; often attributes that are both hidden and subtle. In cataloging solutions for the purpose of such an analysis the authors created two primary and perhaps primitive schemas, one for purpose and the other for mechanism, that would describe any given solution to any specific problem. One could imagine that for any given knowledge domain there might be a host of schematic variables that could be codified and evaluated.

The aspiration behind this work was not to produce consummate existing solutions to unsolved known problems but rather to provide examples of existing solutions that were sufficiently suitable to spark creativity and promote positive ideation. A result the authors believe they have accomplished.

If AI is to get beyond its current phase of assimilating existing data for human interrogation and move to one where it can start connecting some serious dots, this could prove to be a very promising avenue. Areas such as genomic medicine and high energy physics might benefit enormously from the refinement of these techniques. In a few more years we might even get to see a very rudimentary form of synthetic consciousness (see Spooky Action at a Distance – February 2011).

And that might be a singularity worth sticking around for (see Hacking the Future 2.0 – January 2015).

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Insights on Technology and Strategy