Abstract

For most AI practitioners, the unofficial king of questions received is: “Is your model 99% accurate?” Rather than directly answering the question, we shall instead reflect on the following:

• What's the problem definition?
• How do we evaluate the performance of the system? Accuracy?
• What are the costs of a wrong prediction?
• What's the current human performance? Is it reliable or consistent?
• Where does the proposed solution lay in the grand scheme of the process?
• Can we change the problem?

Through a new prism of thought, we will take the audience to examine real-world AI projects developed at Inventec.

Bio.

Trista is a tech executive and AI scientist. She is currently the Chief AI Officer at Inventec, a world-leading tech manufacturer with annual revenues exceeding USD 18 billion. Trista leads Inventec in smart manufacturing, health AI, robotics, MLOps, and mixed reality. Her research not only gets recognized by top AI/CV/medical journals and conferences but has won the world championship in USAID Intelligent Forecasting Competition. Previously, Trista held leadership positions at startups and research labs. At the startup Cognitive Networks, which was successfully acquired, she led the algorithm and data team. At Intel, Trista conducted research on computer vision algorithm-and-hardware co-design. With the Intel team, she also developed the world's most widely adopted CV software, OpenCV, which was downloaded 18 million times as of 2021. At Nvidia, she architected Nvidia's first video processor. Trista received her Ph.D. from Carnegie Mellon University and M.S. and B.S. from National Tsing-Hua University. She co-authored 30+ publications, 70+ issued and pending patents.

All faculty and students are welcome to join.