By Matthew Hutson
The video above will be familiar to anyone who has played the 3D world building game Minecraft. But building those castles, trees and caterpillars is not human – it is a fake experience.
The algorithm takes its idea from the “Game of Life,” a so-called cellular automaton. There, squares in a grid turn black or white over a series of timelines based on how many of their neighbors are black or white. The program mimics biological development, in which cells in an embryo behave according to pupae in their local environment.
Some researchers have replaced the simple rules (e.g., any white square with three black neighbors turning black) with more complex ones determined by neural networks, machine learning algorithms that largely simulates the stringing of the brain. This is called “cloud cellular autata”. But the grid is still only in two dimensions, or in three with just one type of building block.
In a paper posted on the arXiv preprint server this month, researchers installed a system that uses neural neural automata in 3D, and with 50 types of blocks, including some that act as pistons. Then they dismantled their system Minecraft.
The scientists taught cloud networks to grow single cubes into complex designs containing thousands of bricks, such as the castle or tree or furniture building with furniture above, and even into action tools, such as the caterpillar. . And when they cut creation in half, it revived. (Usually included Minecraft, a user had to rebuild the object manually.)
Going forward, the researchers hope to train systems to grow not only in predefined forms, but to design a design that fulfills certain functions. This could involve flying, allowing engineers to find solutions that human designers would not have seen before. People may build these devices in the real world. Or tiny robots could use local interaction (if your neighbor is doing an X, do Y) to collect rescue robots or self-healing buildings. It is, ahem, a growing field.
Source