Built-in Priors
Intelligence may not begin with a blank slate.
Living systems often begin with structures, biases, and expectations that shape what they notice, how they learn, and what they do next.
Principles
What we are learning about intelligence.
Intelligence is not a single mechanism.
Across nature, we see different systems learning, adapting, exploring, remembering, and responding to their surroundings in remarkably different ways.
Project Instinct is an attempt to understand the principles underneath these behaviors.
These are our current working principles of intelligence. They are not rules for how intelligence must work. They are ideas we are testing, refining, and sometimes challenging as the research develops.
The Working Principles of Intelligence
Curiosity
Intelligent systems do not simply wait for information.
They explore. They seek what they do not yet understand and use that exploration to expand what they know.
Active Learning
Learning can be an active process.
An intelligent system can act, observe what happens, and use the result to change what it does next.
Relationships
Information becomes more useful when its relationships are understood.
Intelligence may depend less on isolated pieces of information and more on how things connect to one another.
Interaction
Intelligence develops through contact with the world.
Systems learn not only from information, but from acting within an environment and experiencing the consequences.
Memory
Intelligence depends on what a system can retain and make useful later.
Memory is not simply storage. It shapes what a system recognizes, predicts, and does next.
Forgetting
Forgetting can be part of intelligence.
A system that retains everything may become less useful over time. Intelligent memory may require deciding what no longer matters.
Adaptation
Intelligence changes with circumstances.
A useful behavior in one environment may fail in another. Intelligent systems must be able to adjust as conditions change.
Emergence
Complex intelligence can arise from simpler interactions.
No single component needs to contain the whole behavior for something greater to emerge from the system.
Self-Organization
Order does not always require central control.
Natural systems can organize themselves through local interactions, feedback, and simple rules.
Prediction
Intelligence is partly about anticipating what comes next.
Systems use what they have experienced to form expectations about possible future states.
Abstraction
Intelligence can move beyond individual experiences.
Systems can identify patterns that remain useful across different situations rather than treating every experience as completely new.
Reorganization
Learning does not only add information.
An intelligent system may need to change how existing knowledge is structured when new experience no longer fits what it already knows.
Efficiency
Intelligence does not require unlimited resources.
Natural systems operate under constraints and still achieve remarkable behavior. Those constraints may themselves shape how intelligence develops.
Purposeful Action
Intelligence connects understanding to action.
A system does not merely represent the world. It uses what it knows to choose what to do within it.
Still Evolving
These principles are working ideas, not finished answers.
As Project Instinct develops, some will become stronger, some will change, and others may be added or discarded.
That is the point.
We are not trying to define intelligence before we understand it.
We are trying to understand it well enough to build differently.