"AI Masters Design Utilizing Visual Cues"

2019-11-07

Recently, a study was published in the journal "Mechanical Design" of the American Society of Mechanical Engineers, which states that trained artificial intelligence can use human design thinking to solve problems. The study was co-authored by Jonathan Cagan, a professor of mechanical engineering and interim dean of the College of Engineering at Carnegie Mellon University, Ayush Raina, a doctoral candidate in mechanical engineering at Carnegie Mellon University, and Chris McComb, an assistant professor of engineering design at Pennsylvania State University. Design problems require creative and exploratory thinking. Traditionally, when engineers try to use artificial intelligence to solve such problems, it is mostly applied to a set of well-defined, rule-based problems, rather than following human thinking patterns to create new things. This new study considers an artificial intelligence framework that can learn human design thinking by observing human data, and generate new designs without explicit goal information or guidance. The co-authors wrote in the article, "Artificial intelligence is not just about imitating or replicating existing solutions. It can also learn how humans solve specific types of problems, learn human thinking patterns, and create new design solutions from scratch." The focus of the research is on truss problems, which represent complex engineering design challenges. Trusses are common in bridges and are structures made up of members connected to each other at the ends with hinges. Artificial intelligence observes the progress of design modification sequences through training, which are truss designs created by engineers using the same visual information on the screen. When it's the AI's turn to design, it imagines a design process similar to what humans use and then generates design actions to achieve them. Visualization is particularly important in this process because vision is an important component of how humans perceive the world and solve problems. The framework consists of multiple deep neural networks that work together based on predictions. Artificial intelligence can utilize neural networks to examine a sequence of five consecutive images and predict the next design based on the information collected from these images. Researchers tested multiple artificial intelligences and found that their average performance was better than humans. However, this success is only specific to certain conditions and products. In reality, artificial intelligence still cannot surpass humans because they cannot work without specific goals. In conclusion, the co-authors stated, "People often think that this kind of artificial intelligence will replace engineers, but that is not the case. On the contrary, it can fundamentally change the way engineers work. If we can delegate boring and time-consuming tasks to artificial intelligence, we can free engineers to think boldly and creatively solve problems."

RECENT AI NEWS

RECENT AI TOOLS