In the afternoon on May 31, the 2023 COMPUTEX Forum held its third session, “Chips and Semiconductor – Advance Application Driven by Semiconductor Technology.” The session featured five speakers who each presented their respective company’s latest in edge AI technology.

First up was Sameer Wasson, the VP and General Manager of the Processor Business Unit at Texas Instruments (TI). His talk introduced what makes TI’s AM6x4 vision processors unique and scalable, and how edge AI enhances their effectiveness. Not only did he discuss the applications of edge AI in embedded systems, but also mentioned TI’s mission to make this technology available across industries and consumers. “We are accelerating the adoption of edge AI and making it simpler and more adaptable across a variety of equipment and electronics,” said Wasson.

The next presentation, “Refined Performance at the Edge Computing for IoT and Telecom Open TAN Modern Workload Optimization,” was delivered by Mory Lin, the VP of IoT/Embedded and Edge Computing at Supermicro. Lin highlighted the enhanced performance of Supermicro’s edge computing in devices that are compact, without fans, expandable, and AI-equipped. These devices are like “AI in a box” that can reduce costs and increase efficiency. Furthermore, Lin mentioned how Supermicro invites consultants to help clients find the most optimal business solutions by using edge AI technology.

Lin was followed by Avi Shetty, the Senior Director of Client Strategic Planning and Marketing for Solidigm. He presented Solidigm’s innovative edge AI software for PC data storage and management, Solidigm Synergy. This software helps manage data storage and retrieval for the PC user. “The system knows you and knows your unique behavior, and proactively fetches files before you ask,” Shetty explained. Solidigm Synergy uses edge AI to learn the user’s specific behavior, helping them find files and launch applications more efficiently.

Then came Ali Osman Örs, the Global Director for AI ML Strategy and Technologies at NXP.  He discussed the value of hybrid cloud-connected edge technology, as well as new advances in security for machine learning (ML) software. “What’s valuable is the very hard-to-get dataset that creates a very unique and differentiated machine learning model,” Örs said. NXP’s security strategy is to embed these ML models with a watermark behavior that is easy to detect but does not decrease performance. If this ML model was copied, this watermark would also be copied and easily identified.

Finally, Arnaud Julienne, the VP of Microcontrollers and Digital ICs for the Asia-Pacific Region at STMicroelectronics, took to the stage. In a refreshing contrast, Julienne opened by discussing the applications of edge AI for building a sustainable world. “Let’s take a look at what we are doing to protect the planet while we manufacture tens of billions of chips every year,” he said. He provided two examples of sustainable solutions assisted by edge AI: a smart washer that saves detergent, water, and energy, as well as AI that predicts arc faults in solar panels. He went on to explain the STM32 chip and its edge AI applications.