Elevate your business information technology and strategy at Transform 2021.
This week in a paper printed in the journal Nature, researchers at Google detailed how they used AI to design the next generation of tensor processing units (TPU), the company’s application-specific integrated circuits optimized for AI workloads. While the work was not novel — Google’s been refining the method for the greater part of decades — it gave the clearest case yet of AI’s potential in hardware design. Previous experiments didn’t yield commercially viable products, just prototypes. However, the Nature paper indicates AI can at the least augment human designers to accelerate the brainstorming process.
Beyond processors, companies such as U.S.- and Belgium-based Oqton are applying AI to design domains including additive production. Oqton’s platform automates CNC, metal, and polymer 3D printing and hybrid additive and subtractive workflows, like producing castable jewelry wax. It indicates a range of optimizations and fixes informed by AI inspection calculations, as well as by pre-analyses of component geometry and real time calibration. By way of instance, Oqton can automatically adjust geometries to acquire components within required tolerances, simulating heat therapy effects like warpage, shrinkage, and anxiety relief on titanium, cobalt, chrome, zirconia, and other substances.
While it’s still in the research phases, MIT’s Computer Science and Artificial Intelligence Laboratory developed an AI-powered instrument called LaserFactory that can print completely functional robots and drones. LaserFactory leverages a three-ingredient recipe that allows users produce structural geometry, print traces, and build electronic components such as sensors, circuits, and actuators. As the researchers behind LaserFactory note in a paper describing their job, it might in theory be used for tasks like delivery or search-and-rescue.
At Renault, engineers are leveraging AI-powered software created by Siemens Digital Industries Software to automate the design of automated manual transmission (AMT) systems in automobiles. AMT, which behaves like an automatic transmission but enables drivers to change gears electronically using a push-button, may take as much as a year of trial and error to ideate, build, and completely validate. But Siemen’s tool empowers Renault engineers to manually drag, drop, and connect icons to create a version of an AMT. The program predicts the behaviour and performance of the AMT’s components and creates any necessary refinements early in the growth cycle.
Even Nutella is tapping AI for physical products, employing the technology to pull from a database of dozens of patterns and colors to make different versions of its own packaging. In 2017, working with marketing agency Ogilvy & Mather Italia, the company splashed over 7 million unique layouts on”Nutella Unica” plantations during Italy, which sold out in a month.
People may perceive these applications as taking agency from individual designers, but the coauthors of a current Harvard Business School working paper assert that AI actually empowers designers to overcome past constraints — from scale and scope to learning.
“In the context of AI factories, solutions may even be more user-centered, more creative, and continuously updated through learning iterations that span the entire life cycle of a product. Yet, we found that AI profoundly changes the practice of design,” the coauthors write. “Problem solving tasks, traditionally carried on by designers, are now automated into learning loops that operate without limitations of volume and speed. These loops think in a radically different way than a designer: they address complex problems through very simple tasks, iterated exponentially.”
In a recent blog post, user experience designer Miklos Philips echoed the findings of the Harvard Business Review paper contributors, noting