So maybe its useful to think of a starting point for intelligence as being a group of related microorganisms navigating a surface looking for food and water. The group will conserve energy if it can work together.
Suppose there is a certain path on the surface that leads more directly back to the food or water by going around a few distracting or dangerous areas. So what they really want is to work together to create a map and communicate it to the entire colony.
So by some means they integrate information about their local area, or maybe they can just consult their neighbors each step of the way to find out if they are going the right way. So if the colony is spread out and neighboring the junction points, maybe each one only needs to remember the direction to the next waypoint rather than the full map.
This even seems related to computer vision for 3d reconstruction of a scene, in a way. Because I have been thinking about ways to recognize points, lines, curves, shadows, combine that into surfaces and into shapes. Then recognizing objects based on combinations of shapes.
Say the different types of straight and curved, short and long lines, surfaces and shapes were stored as patterns, similar to the way that the microorganisms supposedly make their map. Maybe its a matter of connecting up a group of units, each of which stores something like a gradient. But the groups can be combined to form higher levels.
Maybe its something like, part of the colony finds food, and then sends a signal out saying "okay, remember this configuration, it leads to food". And then each organism makes some type of record of its own position relative to others in the colony or environmental landmarks, indexed by "food". Or something.
Like a bunch of chains of fuzzy gradients of different dimensions and levels of abstraction, indexed by activity or state.
All of the units index their own subsets of many different pieces of information at the same time.
> Maybe its something like, part of the colony finds food, and then sends a signal out saying "okay, remember this configuration, it leads to food". And then each organism makes some type of record of its own position relative to others in the colony or environmental landmarks, indexed by "food". Or something.
Suppose there is a certain path on the surface that leads more directly back to the food or water by going around a few distracting or dangerous areas. So what they really want is to work together to create a map and communicate it to the entire colony.
So by some means they integrate information about their local area, or maybe they can just consult their neighbors each step of the way to find out if they are going the right way. So if the colony is spread out and neighboring the junction points, maybe each one only needs to remember the direction to the next waypoint rather than the full map.
This even seems related to computer vision for 3d reconstruction of a scene, in a way. Because I have been thinking about ways to recognize points, lines, curves, shadows, combine that into surfaces and into shapes. Then recognizing objects based on combinations of shapes.
Say the different types of straight and curved, short and long lines, surfaces and shapes were stored as patterns, similar to the way that the microorganisms supposedly make their map. Maybe its a matter of connecting up a group of units, each of which stores something like a gradient. But the groups can be combined to form higher levels.
Maybe its something like, part of the colony finds food, and then sends a signal out saying "okay, remember this configuration, it leads to food". And then each organism makes some type of record of its own position relative to others in the colony or environmental landmarks, indexed by "food". Or something.
Like a bunch of chains of fuzzy gradients of different dimensions and levels of abstraction, indexed by activity or state.
All of the units index their own subsets of many different pieces of information at the same time.