[goaty]

For example, a glyph for crocodile which resembles a crocodile is something resembling geometric onomatopoeia, so learning is simplified. Conversely, "to bounce", as written, is further removed; requiring more points of verification, and therefore time, to be integrated by the brain.

When I tried to merge your two examples into a single mental image I came up with bouncing crocodiles. I like that. Sounds like either a child's toy or a rock band.

You beat me to the statement about the drawbacks of time--however much information an object contains, as soon as you move that object continuously in time you suddenly need an uncountably infinite amount of information to describe it. As soon as databases, neural networks, logic, or any other computer constructs attempt to incorporate time, we find the required memory space shoots out of control, as well as the complexity of the description.

As for LOCKSUIT's later comment on logic... There are many types of little-known logics, such as temporal logics, multi-valued logic (https://en.wikipedia.org/wiki/Many-valued_logic), description logics, etc., not just boolean logic. Also, speed and efficiency are crucial to intelligence, and I even include 'efficiency' in my own definition of intelligence, so even though in theory a digital computer or Turing machine can run any algorithm, if the data structure used is not efficient for the application then an answer/response from the computer may not happen fast enough to ever be useful, which by my definition would not be a demonstration of intelligence of any appreciable degree.

If I weren't so busy on my article this week I'd comment more at length on this and other threads.

Ah common u know this one- About that information overload due to trivial circumstances,   you only need 1 cell to fire, but it has to fire in all circumstances - invariance.  And it ends up being a form of compression.  It works feedforward into your robo-brain,  but does it work backwards to dream back out of 1 cell all possible positions that fire the dumbbell?  I bet theres a trick for that.

theres only 1 logic to me,  logic itself, but of course that's just definition arguments.

[AndyGoode]

you only need 1 cell to fire, but it has to fire in all circumstances - invariance.  And it ends up being a form of compression.  It works feedforward into your robo-brain,  but does it work backwards to dream back out of 1 cell all possible positions that fire the dumbbell?  I bet theres a trick for that.

Right on track. That's why people were mentioning object-based reasoning in this thread. If you can develop a system that reliably extracts the most useful invariants from sensory input, especially virtual objects, then you have a good starting point for more advanced logic.  That relates to what I was saying about efficiency: an object-based reasoning system would probably be very efficient because it's working with high-level, compressed, useful abstractions, the same abstractions that people probably use, so that would be a likely starting point for true human-style intelligence/reasoning.

I wish I knew if there were a term for that division between the mentioned hierarchies, and what it is. Maybe 'early vision' or 'low-level vision' versus 'late vision' or 'high-level vision'. Low-level vision is a very challenging problem and doesn't interest me much, so when designing AI systems I like to assume that someone has already solved that problem as a foundation to the more interesting problem of human style reasoning. In practice, however, the two problems are coupled because the brain recalls objects similar to those with attributes similar to the inputted low-level objects and their components, then presumably delivers hypotheses as high-level objects to match against that low-level input. That greatly speeds object recognition, and it's a good system, but it requires solving at least two very difficult problems, one for low-level vision, one for high-level vision.

https://msu.edu/~fcdyer/ZOL867/ZOL867VisionII.pdf