As populations age and need support the market for digital companions for company as well as to assist in daily tasks will explode.
Core technologies to support this include language skills, navigation and mapping, dexterity, skin sensing, muscle coordination, and task learning.
By developing based on synthetic skeletal ligament and muscle fiber systems we can provide a much more human like companion than the TOBOR style robots common today.
These systems are also going to drive advances in limb replacement and exoskeleton assist systems.
Finally development into synthetic brains with billions of neurons to support vision, mapping, thought, emotion, task planning, and muscle coordination is possible with new massively parallel chip designs. These advances will challenge our Turing tests and pondering of what qualifes as conscious and epi-phenomenon in sufficiently advanced systems.
Robotics and Cybertrons as helpers and assistants will drive humanity forward. But we must chose which world we want. Good natured human like or clunky armored TOBORs who can barely walk. Our vision is to create a human form which can do a cartwheel, a tumbling flip, and then catch a pea falling off a table.
This focus on flexibility and agility is not as unimportant as you might consider at first blush. During the Fukushima Nuclear crisis, neither America nor Japan had a robot that could navigate around the wreckage to enter the site.
We are working hard to make this kinder vision of AI and Cybertrons a reality. Using 3D printing of different materials at the same time we can design supporting skeletons in carbon impregnated polycarbonate for strength while adding flexibility using Delrin joint and flexions. It is a very different approach. Rather than using servos for movement new muscle systems based on magnetism will be used.
One critical area is supplying a cognitive center for true cognition not simple object recognition. To do this we must move away from tensor architectures for neural networks to more brain like three dimensional self organizing structures. Based on 20 years of research, Noonean's "neural cube" architecture will provide a next level of cognition for autonomous systems.
These are research areas for now and a lot of technologies have to upscale to make this vision possible. But the vision of more human robotics will be critical for the direction we take.