Interesting, you talk about a method where we look for a small simple lowdata lowprocessing 'complexity' algorithm that pack AGI punch.
Unfortunately, to do R&D, I think you need a lot, lot of data. You can discover facts with little data, it scales!
Complexity theory offers me the thinking tools in which to design and code my type of AI. If the end goal is to identify, monitor and have the many pot holes in my town fixed, my data sets will be location, the width, length and depth of the pot hole. I would be looking to use multiple drones fitted with sensors that could sweep the entire town in one day where I obtain my data in real time of high relevance to my end goal. I could do this repeated times over a period of time to build up a sense of how long it is taking the pot holes to be fixed and how quickly they are getting worse, against local weather conditions and even by monitoring traffic usage of the roads. All these data sets I would not have to pay for, I would obtain them from whatever drones I deploy to monitor a location.
I could divide tasks by many parts, each which could be acting independently but in cooperation with others using feedback loops, with each part powered by its own energy source. If one part fails, other parts can take over the task, the system would continue to function. The sense of oneness or completeness, as in the system has a sense of identity separate from the environment is an emergent state of many parts relating together through feedback loops, which is how nature works, such as bee and ant colonies or our brain.
Complexity theory encourages the designer to think how parts of a system are related, what information is flowing through those connections. A system can be broken up into parts related by function, so a drone only needs to have scripts and obtain information for a small section of location, with multiple drones doing the same thing at different locations. This information can then move up to a different set of parts with a different function of fitting the information together into a holistic whole, doing something to that information, move that information to another set of parts to dump that information out as outputs to the relevant people such as politicians, public, media and those that fix pot holes in real time.
In complexity theory, intelligence is an emergent by-product of the actions of parts, so I am focussed on the parts rather than the emergent layer, which I expect will come about naturally. I can arrange the parts by function, and leverage their relationships and flow of information towards a particular "basin of attraction" that makes them more likely to achieve an end goal.
Each Cymbod I design and deploy is in reality a system of many parts which acts as if it is one thing, rather like ten thousand bees create an emergent state of one entity that is a hive, colony, nest or swarm. The Athorybia rosacea is seen by most people as one entity, the jelly fish, but is really a colony of many smaller parts working in cooperation.
