Can 3D Spatial Memory fix the "Information Retention" problem in AI?

A senior researcher from NCAT has proposed a radical solution to a common AI frustration: the inability to retain information from long, complex conversations with models like GPT-4 or Claude. The problem, dubbed the 'Infinite Scroll' issue, stems from presenting all interactions in a flat, 2D chat window—a format that clashes with how human memory naturally works. Our brains evolved to remember information based on its location in physical space, not as a sequential list. The constant scrolling and lack of visual hierarchy make it difficult to build a coherent 'mental map' of a project, leading users to repeatedly ask the same questions because they can't find past answers.

To address this, the researcher built a prototype called 'Otis' using Three.js and the OpenAI API. Otis transforms the AI interaction into a 3D spatial environment. Instead of a linear chat log, every AI response, code snippet, or piece of research data is saved as an object at a specific 3D coordinate. This spatial anchoring is designed to trigger the brain's powerful spatial memory systems, with research indicating potential retention improvements of up to 400%. The environment offers a persistent state that doesn't refresh like a browser tab, and allows users to create a visual hierarchy—clustering key concepts in the foreground and archiving less critical data in the distance, effectively creating a navigable memory palace for their work.