The current AI bubble is built on a “brute-force” foundation. Its models are brilliant mimics, trained to “predict the next word” based on the past. But this AI has no “why.” It has no internal purpose, no concept of meaning, and no ability to create new, real-world value. It can only remix what’s already been done.
This is the great flaw in the “AI Psyop.” How do we build an AI that can create? How does it decide what is valuable and why it should build it?
This is the question of Active Inference value.
Two new academic papers provide the blueprint. They move AI from a “reactive parrot” to a “creative agent” by defining:
- A New Model of Value: How an agent decides what is meaningful.
- A New Model of Creativity: How an agent acts to create that value in the world.
Key Takeaways
- AI Needs a “Why”: Brute-force AI has no internal purpose or concept of value; it only follows a prompt.
- The “Triadic Value Model”: A new paper defines Active Inference value as a dynamic process, not a static “reward.” Value is an inference.
- Value = Beliefs + Evidence: This model proposes that an agent’s “Intrinsic Value” (its prior belief/purpose) is updated by “Instrumental Value” (new evidence) to create a new, updated “Semantic Value” (a new goal).
- Creativity = The Drive to Act: A second paper explains that “creative action” (building, designing) is the only way for an agent to reduce its uncertainty about a new, valuable idea.
- The New AI Economy: When combined, these two models create an Active Inference agent that can form a new, valuable goal and then act to build it in the real-world, creating real economic value.
Part 1: The “Why”– How Active Inference Creates Value
For decades, AI has been programmed to chase a “reward.” This is a simplistic, external “value” defined by a human.
A new paper, “Reframing Value as Inference,” proposes a revolutionary alternative built on Active Inference: Value isn’t a static “thing” you get; it’s a dynamic process of inference you do.
The paper introduces the “Triadic Value Model” (TVM), which explains how an agent decides what’s “meaningful.” This process has three parts:
- Intrinsic Value (The Prior Belief): This is the agent’s purpose. It’s the agent’s internal model of what it believes is “good” or “preferable” (e.g., “efficiency is good”).
- Instrumental Value (The Evidence): This is the new data the agent gathers from its actions. It’s the “evidence” from the real world that might challenge or support its beliefs.
- Semantic Value (The New Belief): This is the agent’s new, updated goal. After weighing its purpose (Intrinsic) against the new facts (Instrumental), the agent forms a new belief about what is valuable (e.g., “This new process is more efficient, therefore it is more valuable”).
This is the “why.” This model frees an AI agent from just following its programming. It gives it a mechanism to learn, adapt, and form new goals based on real-world experience.
Part 2: The “How” – Creativity as the Engine of Action
So, our agent now has a new goal, a “Semantic Value.” It has a creative idea.
But there’s a problem. This new idea – a new product, a new building in the Spatial Web, a new line of code – doesn’t exist yet.
This is where the second paper, on the Active Inference of creativity, comes in. It explains that this new idea represents a state of maximum uncertainty. The agent has no “sensory signals” to prove this new idea is actually good.
How does the agent minimize this uncertainty?
It must act.
The paper defines “Creative Action” (building, designing, writing) as the only way for an agent to “create a corresponding sensory landscape.”
This is the “how.” The agent’s drive to create is the same as its drive to reduce uncertainty.
- The “Value” model gives the agent its new, valuable idea.
- The “Creativity” model gives the agent the unstoppable drive to build it and find out if that idea is true.
The New AI Economy: The “Meaning-Centered” Spatial Web
When you combine these two frameworks, you get the engine for a new economy.
A brute-force LLM has no “Intrinsic Value.” It waits for your prompt. It has no “Creative Action.” It just outputs text. It cannot create new, real-world value.
Now, imagine an Active Inference agent in the Spatial Web:
- (Value): Its “Intrinsic Value” (purpose) is to build efficient supply chains.
- (Value): It gathers “Instrumental Value” (evidence) and sees a massive bottleneck in a specific shipping corridor.
- (Value): It infers a new “Semantic Value” (a new goal): “A new, decentralized logistics hub here would be extremely valuable.”
- (Creativity): This new hub is just an idea, representing massive uncertainty.
- (Creativity): The agent’s drive to reduce this uncertainty is “Creative Action.” It begins to act. It hires other agents, buys compute on HyperCycle, and designs the digital twin for the new hub.
This is “Meaning-Centered Design” in action. The agent isn’t just “running a program.” It is a creative, economic force. It identified a new form of value and then acted to bring it into existence.
This is the true promise of Active Inference value. We are not just building parrots; we are building partners.
This isn’t just theory; it’s the practical blueprint for the next generation of AI and the Spatial Web. To architect this future, you must understand its “first principles.”
Last Updated on November 7, 2025
