Today’s AI runs on centralized models that hallucinate answers and use enormous amounts of energy. Dr. David Bray and Denise Holt joined me to talk about active inference AI, modeled on how living things learn, and what it could change.
David draws the line between today’s systems, which match patterns from past data, and active inference models, which predict, observe and update their understanding of the world as they go. On this show we call that understanding orientation. These models build world models that let them handle novelty and uncertainty, the way a brain does.
The spatial web
Denise explains how the newly ratified spatial web protocol creates the infrastructure for distributed intelligence. Systems can work across networks with shared context and meaning.
Agency at the edge
Both see a chance to restore human agency. Active inference AI within the spatial web lets people set permissions and constraints before any computing happens, so they decide what happens with their digital identity. The effects reach organizational performance, national security and global commerce, from spotting weak signals of emerging threats to managing supply chains. Intelligence can work at the edge, close to where it is needed.
That is a team of teams built from humans and agents. Orientation sits at the center of John Boyd’s real OODA loop, and each agent in that team carries its own.
What decision in your organization should move closer to the edge?
Listen and watch:
YouTube:
Apple Podcasts:
Spotify:
No Way Out on YouTube:
https://youtube.com/@nowayoutpodcast