The findings published in the ASME Journal of Mechanical Design reveal that AI agents can solve engineering problems by adopting human design strategies.
Designing is a skill requiring creativity and a problem-solving ability which humans possess. Now, although artificial intelligence (AI) has already been used in designing, it has been just applied to problems within a set of rules defined by humans because AI is still unable to create new designs by its own. However, this novel research intended to train AI to learn human design strategies by observing the humans’ process of modification in creating a complex structure, so that AI can design without humans’ help.
This research was conducted by Jonathan Cagan, professor of mechanical engineering and interim dean of the college of Engineering; Ayush Raina, a Ph.D. candidate in mechanical engineering at CMU; and Chris McComb, an assistant professor of engineering design at the Pennsylvania State University.
“The AI is not just mimicking or regurgitating solutions that already exist,” said Cagan. “It’s learning how people solve a specific type of problem and creating new design solutions from scratch.” How good can AI be? “The answer is quite good.”
In the research, “a truss” is the major subject used to train AI agents as it represents complex structure. Using the same visual information that engineers use—pixels on a screen, researchers trained AI agents to observe how a truss structure was created, designed and modified step by step. Incredibly, when AI agent started to design, it performed in the same way as how humans do; AI agents imagined design progressions, and generated design moves to realize them. The researchers defined “Visualization in the process” as the key since vision is the ability to identify and solve the problems.
The researchers carried on the research within the framework of multiple deep neural networks that served in predicting the possible situation. With this framework, AI agents understood what the next design should be after scanning a set of five sequential images.
“We were trying to have the agents create designs similar to how humans do it, imitating the process they use: how they look at the design, how they take the next action, and then create a new design, step by step,” said Raina.
Also, AI agents were tested on similar problems, and the findings showed that AI agents performed more effectively than humans. However, it did not show many advantages like what humans offer. That means AI agents weren’t design with a specific purpose, such as making something lightweight, and did not receive feedback on how well they were doing. In contrast, they just designed by using the vision-based human strategy techniques that they had been trained for.
“It’s tempting to think that this AI will replace engineers, but that’s simply not true,” said McComb. “Instead, it can fundamentally change how engineers work. If we can offload boring, time-consuming tasks to an AI, like we did in the work, then we free engineers up to think big and solve problems creatively.”
This paper is part of a larger research project sponsored by the Defense Advanced Research Projects Agency (DARPA) about the role of AI in human/computer hybrid teams, specifically how humans and AI can work together.
.jpeg)
