edfas.org 39 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 26 NO. 1 focused topics. Enthusiastic engagement characterized all six user groups, with attendees passionately discussing pivotal issues pertinent to their roles as analysts and generously sharing their wealth of knowledge and experiences. Diverse participants, including analysts representing major semiconductor companies, tool vendors, and esteemed universities, contributed to the vibrant The primary goal of the newest User Group at ISTFA was to arouse curiosity and engage participants in emerging AI topics. The chosen focus was the imperative need for standardization in various aspects and components of IT and AI systems. Only through standardization can the entire industry harness the benefits of AI applications. The pedagogical approach aimed to foster participant engagement by alternating between teaser contributions and explicit questions posed to the audience. The initial contribution provided an overview of multiple domains of AI applications in failure analysis. An accompanying graphic visually depicted these domains, ranging from defect classification and neural network applications within tools, to pattern recognition across samples and tools. Additionally, it highlighted AI applications in failure analysis reports, classification, and cross-referencing to previous similar cases in semiconductor companies. Following this, a teaser focused on an actual application in acoustic microscopy. Unlike commonly displayed image segmentation and classification, it showcased the classification of acoustic microscopy A-scan signals. The phase and amplitude information of an A-scan represented the color of the respective pixel in a C-scan. These A-scan signals were digitized for each C-scan pixel, and the resulting vectors were fed into a neural network for training, requiring substantially fewer training “images.” The use of A-scan vectors instead of C-scan images was slightly above the audience’s understanding level. However, this led to numerous questions and a fruit- ful discussion. The session concluded with a video demonstrating AI-enabled failure analysis workflows. The video depicted exchange. The overwhelming feedback from attendees lauded the sessions’ content, format, and the high level of engagement observed throughout. Encouragement was extended to the audience to prolong their conversations via dedicated ASM community discussion boards, fostering a continuous and collaborative exchange of insights. ISTFA 2023 ARTIFICIAL INTELLIGENCE (AI) IN FAILURE ANALYSIS USER GROUP Chair/Co-Chairs: Florian Felux (Infineon Technologies AG), Thomas Rodgers (Zeiss), and James Demarest (IBM) florian.felux@infineon.com, thomas.rodgers@zeiss.com, jjdemar@us.ibm.com the standard proposal of sample convergence, physical sample holder, analysis tool results, and respective metadata in JSON text format. Although discussion materials and explicit questions were prepared, they were not utilized as the spontaneous discussion filled the scheduled one-hour timeframe. The main contributions from the audience in the discussion revolved around the trustworthiness of AI applications, data and model sharing, and training aspects for FA analysts. Particularly, participants were keen to understand how to initiate hands-on first steps in AI applications. A notable comment at the session’s end was an audience member’s query: “Doesn’t that necessitate standardization?” In response, the User Group chairs suggested continuing the AI user group as a series of online sessions. The initial sessions could include a hands-on exercise titled “My First AI Application.” Presenters and co-chairs of the new Artificial Intelligence User Group.
RkJQdWJsaXNoZXIy MTYyMzk3NQ==