,
11 tweets,
3 min read
Read on Twitter
It is essential to frame Complex Adaptive Systems from the lens of a learning system. The dynamism of complex systems are a consequence of its learning algorithm. Natural Evolution and Cultural evolution are both learning algorithms.
Evolution consists of generative models that not only generates learning systems but also learns to improve its generative models. We find 2nd order feedback not on just in the system itself but in the generative model of the system.
Perhaps the only pragmatic way to understand a complex system is to understand the dynamics of the generative model. Focusing on the generated model is like focusing on just the trajectory of a Lorenz attractor but ignoring the generating function.
But you still need to go one step higher, what is that object that characterizes functions that generate strange attractors? This object the 'generative model' that I refer to. There are computational models that generate behavior and there are meta-level generative models.
These meta-level generative models are not static but dynamic in the sense of a learning system. The perplexing question is how do systems that do learn propagate their knowledge to the meta-learning level?
To conclude, understanding of complexity requires understanding of complex learning systems. This is why the connectionist paradigm will evolve to an emergent level where learning systems and complexity science will cross-fertilize.
This blog wiringthebrain.com/2019/09/beyond… by @WiringTheBrain spurred this thought stream. Mitchell points in the right direction about reductionism and complexity sciences, but you can sense a great filter in further understanding.
@WiringTheBrain The mechanisms in a complex adaptive system are impenetrable. However, one may have better odds in formulating a generative model. These models may not generate the same exact system, instead, it exhibits similar behavior.
@WiringTheBrain So it's like generating an artificial human mind. It's not the same thing as a biological human mind but it exhibits the same behavior. The same method applies to simulating biological systems.
@WiringTheBrain I conjecture that simulating biological systems is equally as difficult as simulating an artificial mind. There is sufficient complexity in biology that goes beyond how a mind is created. This is actually a harder problem than AGI. Expect AGI to come first before!
@WiringTheBrain @threadreaderapp unroll please
