If you allocate x total Watts for I intelligence required there's going to be lower complexity models within a computational complexity budget. Reduced model complexity is going to cluster around particular operations with power and model complexity varying with efficiency. Self-improving efficiency algorithms with implicit mathematical representations I think are self-organizing. Assuming models are in executable form... that self-organizing structure has to be memory constrained but it could bloat out of control like junk DNA... if the constraint structure is initially formal the bloat can be reduced since junk DNA seems to be a discovered resultant molecular agglomeration accumulated from agents functioning in the environment. So increasing efficiency is an optimizing of the formalizing of the self-organizing, or a malleating. This malleating structure optimally performs symmetry and cyclicity compression whereas non-symmetry/non-cyclicity creates bloat... basically IMO you want the malleating to be a tempered abstract mathematical structure compression while maintaining targeted efficiency in internal communication complexity and query complexity. That tempering constraint needs to be formal as well. But IMO the process doesn't mean explicitly doing full K complexity estimations due to query complexity inefficiencies.
------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Td69923751482264f-M476afa941feb2d5d6f0f1359 Delivery options: https://agi.topicbox.com/groups/agi/subscription
