Thinking leaves a mark in your neural mesh a.k.a. brain

Question was which GPU is still viable to train upcoming AI, its allway beyond personal budget to even rent at 3$/hr as salaries are very constrained, due to lower purchasing power. One would train plenty of datasets, if they dont train faster, they still take a day or two, due to poor learning way, (ideally these systems shuld be capable of updating and learning in real time, up to the point of thinking, which is real time ongoing constant learning, thinking affects the mind, it writes its own history to the network, where thought is being processed.So when these systems in network can evolve continuously by themselves from thinking i.e. constant learning, one network can be termed as one entity, and together collectively they can think.Usual training on all data is in a way thinking, where it tries to learn the entire structure of knowledge, and in the process, takes out common things and patterns that allow it to learn more, as if all knowledge when represented, in network should pass through lower levels of network, which encode most basic structure, and as it propagates, it forms associations which is everything in thinking, and these associations as it progresses can be made act upon itself to produce next states, based on its memory, itself modifying its state to new states, hence thinking.