From smart plants to algorithms
Up until now, artificial intelligence (AI) has been modeled on human intelligence, resulting in software programs that can autonomously solve problems as well as or better than humans. But AI would be more resilient if it were based on the natural life cycles found in the world around us.
As the natural world goes through cyclical patterns — the water cycle, plants’ seasonal growth, etc. — so does the man-made world, with cycles in financial markets or scientific fields. To better survive the peaks and valleys of dynamic fields (like technological progress), why couldn’t systems be improved through biomimicry? The idea is to find sustainable solutions to human challenges by emulating time-tested patterns and strategies modeled from complex infrastructures in the natural world.
For instance, while at first glance a bean and a deep learning model have little in common, the former could serve as inspiration for the latter. After all, plants and their unique intelligence are the planet’s own evolutionary technologies. A durable learning program would be constructed with a seed module, with instructions concerning the end task; a root module, siphoning information towards the stem module and a “solution” leaf module. The system could self-prune, and generate seeds with improved information. Beyond AI, the natural world could influence other systems, from social interactions to marketing and production.
To go further
“Deep Learning with Biomimicry“
by Emily Mendez and Emily Marcum (MIT Journal of Design and Science, 19 February 2019).
© Copyright Business Digest - All rights reserved
