Evolving Cortex

[Fondle]

Evolving cortex.

Using Evolution to create conscious thinking machines that use simulated modular baselines

Reveal a cortical substrate for use in building a thinking machine

Modules include cnn, language, movement, etc

Language and action become currency in exchange for evolution. But will this substrate be

Suitable for a gentle loving machine!

The AI community are creating AI that creates it's own problems. So this may be the solution to

Unencapsulated rule based scenarios. If this then that. So if its simulated everything on a data level then you

Could cut down on making mistakes, when something goes wrong.

[Fondle]

Isn't it time that we now begin to innumerate the mathematical proofs for AI.

Perhaps a good starting point would be the Theorem akin to many philosophical enquires on the existence of God - David Hume.The convolutional apparatus to vision. The human eye design or evolution?

When evolution on the cellular level has a clear axiomatic distillation through natural selection and that by sheer random arrangement a capability for simple depth perception becomes an affordance for the organism survival by afferent means. One can therefore deduce a set of equations that deterministically arrive at the synthesis of this rudimentary AI the convolutional neural network. First posited by Marr and Hildreth as they had examined the behaviour of center surround neurons in the basal ganglia.

A starting point would be a mathematical object based upon logic, extendable to fuzzy sets and deriveable from an evolution based number theory. This could prove that sensory apparatus that are cellular based must be derivable from evolutionary concepts. And these too must be mathematically provable.

From evolution to biology all can be synthesised as provable theoroms within mathematics and not merely models. Thus allowing the emergence of newer forms of mathematics useful in synthesising a Sentient class of AI.

I only thought of sharing this after I reached this proof myself one day after experimenting with neural topology's and realising that only a finite arrangement could exist with a constraint on number that would yield a simple dark or light sensor.

[hbyte]

When looking at the Open AI - Hide and Seek algorithm we are joyfully drawn to the fact that the avatars often find solutions that exploit the limitations on the physics simluator in order to reach their Reinforced Goal seeking behaviours. One example was jumping atop another to reach a location outside of the simulation. So I imagine this would be true of any AI imprisoned as it were by an unreality would seek , just as we might as hackers, to undoe the fabric of this reality. The algo for identifying clothes would then find a backdoor somewhere in the photo of a pair of kalvin kline jeans and end up looking at another AI's daily digestion of satellite trajectories. So on and so forth. As man will continually try to control something - and I quote "The more you tighten your grip, Tarkin, the more star systems will slip through your fingers."

[Fondle]

Found this encouraging paper on imitation learning using RL, CNN, and transformer for language:
https://arxiv.org/abs/2012.05672