Controlling a complex body is pretty much solved.
Here-s a virtual example after 50 days (and ~80G simulation timesteps) of training , and here-s a real-life incarnation of a … hybrid creature.
Sure one can ponder about whether to re-invent body control with SDR/HTM means, or do only body-riding an existing, already trained body, or pick the hard path and do both.
In regard to
I think development of autonomous robot AI will get a boost from war (unfortunately). The reasons are the high stakes and a much relaxed demand on failures. Take a rudimentary (from AI pov) precision shell like excalibur.
it is a $70k expendable piece of equipment. Which is a more expensive than a Tesla model 3, single use “vehicle”
for which a relatively high failure rate of 5% or 10% is just a problem of accounting for the extra costs.
and whenever it accidentally hurts or kills innocent bystanders, it can be excused as an unwanted yet unavoidable collateral damage
Low level machine control. Ports and memory, servos and sensors, timers, drivers, the bits and bytes stuff. Still software engineering, well understood. [spinal cord, brain stem?]
Trainable stuff. Machine learning, zillions of training cases, parameters. Novel, rapid progress but has limits. Nice link. [cerebellum?]
AGI. Predictive, planning, sequencing, passage of time, theory of mind, models, maps. [cortex?]
We’re still a fair way off being to do any of the last stuff. 20 years?
I would wager most of these are rather our minds interpretation for a simpler process available in a simpler form to insects too.
And the main difference between us and insects is more of the same processing units, with the evolutionary breakthrough of an algorithm capable of orchestrating millions of tiny brains without ending in cacophony.
I think it short-sighted to discount the intelligence of so-called ‘tiny’ brains, or even no brain at all like that of slime mold–they both exhibit (unconventional) intelligence. I assert, that the underlying rules of cognition exhibited by simple organisms are the same (in principle) as the neocortex, and a plausible hint as to how the mind emerges from the brain—or the lack of one; that such theoretically based principles are key to discovering (general) intelligence.
I’m building such things in a game called ‘Cyborg’. Not to build a game, but to discover cybernetic patterns and behavior. Find me at: