This sounds interesting. I will like someone from the field of genetic programming on how this works and how it differs from current DL approaches.
His basic strategy is to have a scalable problem decomposition strategy.
So programs that process pixels and the teaming of those programs are grouped together. The groupings (teams) themselves are co-evolved with the programs, simultaneously.
This enables niching and specialization behavior.
This builds on earlier work on 'symbiotic bid-based genetic programming' from other people at Dalhousie, the same university Kelly is at.
The innovation of this paper is that teams can reference other teams.
This allows for the creation of hierarchical teams. (There are rules to prevent cycles and other edge cases.)
Everyone commenting here is probably going to just look at numerical game score and ignore the fact that the runtime performance of Kelly's tangled program graphs. They are 1000 times smaller than a deep neural network. That matters for things like running on mobile/embedded devices.
Ding ding ding. This is where the money is at, good yet cheap sensors that sense human level actions are needed for IoT to be impactful.
I am wondering, whether a similar approach is possible with current DL models and will it have any performance improvements over what is existing or whether it will be computationally even more expensive.