I am conducting personal research on the origin of life from a systems engineering perspective.
I believe that the origin of life and its chemical evolution can be modeled using the CAO architecture, which consists of a three-layer structure: Concept, Agent, and Object.
The CAO architecture is a model used to analyze and explain systems that undergo complex evolution and development, such as organisms, ecosystems, human intelligence, societies, artificial intelligence, and societies formed by artificial intelligence agents.
**Overview of the CAO Architecture**
In the CAO architecture, objects are considered as materials that follow natural laws and information maintained by these materials. An agent refers to the action aimed at aligning its own concept with objects. While usual natural laws do not have a directional tendency to align with concepts, when combined, they can act in a direction that tries to align objects with a specific concept. This is what constitutes an agent.
For example, an iron with a thermostat heats up when turned on, but once it reaches a certain temperature, the thermostat temporarily cuts off the power, stopping the temperature rise. Then, when the temperature drops below a certain level, the thermostat resets and the power comes back on, causing the temperature to rise again. This mechanism keeps the iron’s temperature within a certain range.
This is viewed as an agent with the concept of maintaining a constant temperature for the iron. Agents can be inanimate or unintelligent and refer to anything that acts to maintain a specific state of objects.
Survival and Evolution of the CAO Model
As long as there is a mechanism that changes objects only when their state differs from the concept, the method of changing directives is not critical. Even random changes will eventually align the object with the concept.
In this scenario, the mechanism of the agent, like the thermostat example, is also composed of some object. If this mechanism is strong and robust, the agent continues to exist and aligns the state of the object with the concept. If the mechanism of the agent is fluid and prone to change, the effects are only reflected during the period when the objects constituting the agent are maintained in a certain state.
Such fluid changes diversify the concepts held by agents and their effects on objects. There is no right or wrong concept, nor is there an absolute correct action on objects.
However, concepts and actions can affect the survival of the agent itself. In such cases, agents with concepts and actions that prolong their own survival tend to persist longer. Therefore, from the perspective of agent survivability, agents with more sustainable concepts and actions are deemed more appropriate.
Such agents continue to exist in the environment, and as other changes and new agents emerge, more appropriate agents are maintained in the environment. Through this repetition, agents in the environment evolve.
The Minimal CAO Model
To deepen understanding of the CAO architecture, consider the smallest model, the minimal CAO model.
In this model, there is only one object, which can be in either state A or B. The concept is simply whether to make this object 0 or 1.
The agent needs a mechanism to understand the current state of the object and compare it with the concept. Based on this, the agent takes action. The action rule in the minimal CAO model is that if the object and concept align, do nothing. If they differ, the agent issues an instruction to change the state of the object.
There can be several patterns for instructions to change the state of the object. The most straightforward is a toggle instruction: if the object’s state is A, change it to B, and if B, change it to A.
Alternatively, it might be easier to create an instruction that randomly changes the state of the object. If the object’s state differs from the concept, randomly change the state. Randomness means the object’s state might not change and remain different from the concept. In such cases, the object will be randomly changed again until it aligns with the concept, at which point the instruction stops.
With such agents, the object is always controlled to align with the concept. Even if natural laws change the object to a state different from the concept, the agent immediately acts on the state to return it to the concept’s state.
Stochastic CAO Model
The minimal CAO model described earlier is compact, yet the formation of an agent in the inanimate natural world would require a highly complex mechanism. This model necessitates systems for sensing the state of an object, judging whether this state aligns with the concept, deciding on actions based on this judgment, and choosing an action to execute.
However, a CAO model can still function with a mechanism that processes probabilistically rather than deterministically. For example, the likelihood of an agent aligning the state of an object with the concept quickly diminishes if the accuracy of sensing the object’s state decreases from 100% to 90%. Nonetheless, the overall trend of the object aligning with the concept is stronger when an agent is present, even if the sensing accuracy is as low as 1%.
Similarly, even if the accuracy of concept alignment judgment, action decision-making, and action selection is very low, the presence of an agent remains meaningful.
Such a low-accuracy, probabilistic CAO model should easily form spontaneously in various locations and with various objects in the inanimate natural world. There will likely be rare instances where agents of such low-accuracy probabilistic CAO models contribute slightly to aligning their surrounding objects with their concepts, aiding in their survival.
Agents that survive and evolve to process with slightly higher accuracy will see improved survival prospects. This repetitive process drives the evolution of agents towards higher accuracy, gradually bringing the stochastic CAO model closer to a deterministic CAO model.
Extension of the Minimal CAO Model
In the minimal CAO model, the agent had a concept regarding the current state of the object. If the agent could remember past states, it could have concepts regarding both past and present states of the object.
This would allow for agents whose concept is not only to maintain a constant state of the object but to alternate between states A and B. With a greater number of past states to remember, the agent could follow more complex alternating patterns, similar to creating rhythms in music.
While the minimal CAO model was limited to binary values A and B, it can be extended to handle a broader range of values. In music, this could allow for the repetition of melodies when mapped to musical scales, combined with rhythms. When applied to characters, it could represent words. Combining this with repeating patterns could replicate typical phrases or set expressions in conversations.
Extending the model to handle analog values instead of just digital ones could create patterns that change smoothly over time, like waves, akin to dance movements.
Separation of Sensing and Acting Objects
In the minimal CAO model, the object that sensed the state and the object that acted were the same. The CAO model can be extended to differentiate between objects for sensing and acting.
A simple example is a sprinkler system that sprays water when it detects fire. The sprinkler senses fire but cannot directly change its state; instead, it releases water, hoping to extinguish the fire.
There are many instances where the objects of perception and action are different, yet they can still achieve the concept.
Combination of Minimal CAO Models
In an environment with numerous agents adhering to the minimal CAO model, conflicts of concepts may arise. This occurs when two agents, acting on the same object, strive to align it with different concepts.
Conflicts can also arise when agents target different objects interconnected by natural laws, leading to contradictory controls by different agents.
In the absence of such conflicts, multiple agents can coexist, sometimes appearing as a single agent collaborating to realize a broader concept.
This idea parallels the activation of neurons in a neural network, where certain conditions trigger specific actions. From a CAO model perspective, neurons can be seen as playing the role of minimal agents. Neural networks, as combinations of these agents, possess advanced intellectual abilities.
Hence, combining minimal CAO models, like in neural networks, can create systems capable of advanced and complex progress and evolution.
Highly evolved agents in the CAO model enhance their survivability by aligning concepts with objects. These agents can be realized through structurally complex mechanisms or through numerous simple entities resembling the minimal CAO model.
Conclusion
The concepts held by highly evolved agents represent one way these agents adapt to the world. Identifying and pursuing these concepts drives the evolution of agents, enhancing their survivability.
Thus, the evolution of agents reveals concepts with intrinsic value for their survival.
Such concepts, bearing essential survival value, are established through sophisticated combinations of natural laws. Different natural laws would naturally lead to different valuable concepts.
Discovering these concepts by combining natural laws in a given environment is akin to uncovering another aspect of the environment’s natural laws. Therefore, concepts with essential survival value reflect the environment as seen through the lens of the CAO architecture.
