Eye on AI - July 24th, 2020
Welcome to Aigora's "Eye on AI" series, where we round up exciting news at the intersection of consumer science and artificial intelligence!
This week, we’ll be looking at the influence of nature on AI systems, and how many complex neural networks owe much of their success to the observation and replication of common natural phenomena.
New Network Opens Door for Complex Physical Phenomena Simulation
We begin with a fascinating article out of Science Magazine, titled “Watch artificial intelligence learn to simulate sloppy mixtures of water, sand, and ‘goop’,” in which reporter Matthew Hutson describes how researchers at DeepMind, Google’s AI branch, recently revealed they have trained a relatively new type of AI model called a Graph Network-based Simulator (GNS) to simulate complex physical phenomena by watching natural physics in action.
“The program can realistically recreate the interactions between tens of thousands of particles of different materials, lasting thousands of animation frames,” writes Hutson. “.... The system uses ‘graph networks,’ representing a scene as a network of interacting particles (each particle much bigger than a molecule—some clips above were later rendered in hi-resolution) that pass “messages” to each other about their positions, velocities, and material properties…. Once trained, the system can generalize to never-before-seen situations—predicting the behavior of many times more particles, or what would happen if you added more obstacles like ramps, or shook up the box.”
While predictions aren’t 100% accurate, (they are pretty darn close) the potential application for GNS is practically neverending, from chemical mixture to storm simulations to projection models. That’s big picture. Graph Network applications are still relatively new and need time to scale. Still, the future looks bright. To show how a GNS works, check out this short video of GNS prediction vs. reality.
Sustainable Farming Drones Go to the Birds for Inspiration
Let’s continue with an article out of Loughborough University, titled “Artificial intelligence and food security: swarm intelligence of AgriTech drones for smart AgriFood operations,” in which researchers Konstantina Spanaki, Erisa Karafili, Uthayasankar Sivarajah, and Stella Despoudi Zahir Irani describe how the idea of sustainable farming led them to study how Agricultural Technology (AgriTech) drones navigating in similar biomimetic ways to bird swarms might positively impact farming operations in inaccessible land.
“The study adopts a Design Science methodology and proposes Artificial Intelligence (AI) techniques as a solution to food security problems,” note the research team. “Specifically, the proposed artefact presents the collective use of Agricultural Technology (AgriTech) drones inspired by the biomimetic ways of bird swarms. The design (artefact) appears here as a solution for supporting farming operations in inaccessible land, so as unmanned aerial devices contribute and improve the productivity of farming areas with limited capacity.”
The idea of mimicking animals is nothing new in aerial design (think airplanes and birds, or AI gliders utilizing air currents like birds). What is new(ish) is mimicking animal navigation / communication methods neural networks. We reported last month that ant colonies were routinely being mimicked by AI networks––in fact, AI scientists have been relocating ant colonies for years. Bird flight pattern replication is less common. If the study proves successful, new AI-driven agrifarming practices (and more remote farming land) could emerge.
By continuing to be inspired by the natural world, I’m confident AI researchers will continue to establish newer, more effective means of artificial navigation and communication. For more on how nature is helping AI evolve, check out this article on AI and nature, or this Google blog post about nature-inspired robot agility.
That's it for now. If you'd like to receive email updates from Aigora, including weekly video recaps of our blog activity, click on the button below to join our email list. Thanks for stopping by!