Next Meeting

The 36-Inch Blind Spot: Why Major AI Models Fail at Spatial Reasonin

Abstract

In the rush to build increasingly massive AI models, an invisible architectural blind spot has emerged: spatial reasoning. What happens when you ask the world's most advanced AI engines - from multi-billion-dollar cloud systems to state-of-the-art local Vision-Language Models (VLMs) - to calculate the total height of a simple shelving unit with overlapping sections?

They all fail.

This session explores the fundamental gap between human visual intuition and AI mathematical calculation. Using a straightforward woodworking diagram as a stress test, we will break down why leading models consistently misinterpret basic geometric relationships.

Doug Drechsel

Doug Drechsel has worked on many server side platforms including Tuxedo and WebLogic Server. He is currently a developer advocate working with the Database team and reaching out to full stack developers. His most recent work has been with creating servers as a service using Node.JS, specifically in the Mobile/Web use cases.