To paraphrase Arthur Schopenhauer, genius is seeing what everyone else sees and thinking what no one else has thought. Put another way, genius is breaking down the usual silos that isolate ideas and knowledge into specific fields and purviews. Thanks to work about to be presented by researchers at Carnegie Mellon University's School of Computer Science and the Hebrew University of Jerusalem, it may soon apply to AI. The researchers have just given computers the capacity to mine patent databases and other research records in order to repurpose old ideas to solve new problems. To do it, they had to devise a method to teach computers to make analogies.
A great analogy can often be the key to innovation, making it possible to transfer knowledge from one domain to another. Now researchers have shown that rather than relying on eureka moments, crowdsourcing and AI can dramatically speed up the search for these parallels. Examples of analogies leading to major breakthroughs range from Edison's early work in motion pictures to Kepler's elucidation of the laws of planetary motion. But being able to strip away superficial differences and understand the underlying similarities between solutions to diverse problems has so far largely relied on individual genius. That doesn't need to be the case, though, according to the authors of a recent paper in PNAS.
A couple years ago I wrote about an effort to emulate analogous thinking in order to tackle difficult problems and drive innovation. The research, conducted by computer scientists at Carnegie Mellon University and Hebrew University in Jerusalem, sought to harness the power of a deceptively simple intellectual trick: Solving a complex problem by making unlikely connections to known problem-solution sets that might seem unrelated on the surface. If you could train AI to "think" analogously, the premise holds, you'll drive an explosion of innovation. The researchers are back, abetted by scientists from the Bosch Research and Technology Center in Pittsburgh, the University of Maryland, and New York University Stern School of Business, with a new report published online this week by the Proceedings of the National Academy of Sciences. The report describes a process for utilizing remote workers and AI in concert to identify analogies.
The availability of large idea repositories (e.g., the U.S. patent database) could significantly accelerate innovation and discovery by providing people with inspiration from solutions to analogous problems. However, finding useful analogies in these large, messy, real-world repositories remains a persistent challenge for either human or automated methods. Previous approaches include costly hand-created databases that have high relational structure (e.g., predicate calculus representations) but are very sparse. Simpler machine-learning/information-retrieval similarity metrics can scale to large, natural-language datasets, but struggle to account for structural similarity, which is central to analogy. In this paper we explore the viability and value of learning simpler structural representations, specifically, "problem schemas", which specify the purpose of a product and the mechanisms by which it achieves that purpose. Our approach combines crowdsourcing and recurrent neural networks to extract purpose and mechanism vector representations from product descriptions. We demonstrate that these learned vectors allow us to find analogies with higher precision and recall than traditional information-retrieval methods. In an ideation experiment, analogies retrieved by our models significantly increased people's likelihood of generating creative ideas compared to analogies retrieved by traditional methods. Our results suggest a promising approach to enabling computational analogy at scale is to learn and leverage weaker structural representations.