Once you can grasp Moore’s (very valid) Law and see how all this will accelerate the development of first super-human AI and then the technologies designed by these super-human intelligences to a point beyond imagination, it is quite easy to believe that all of what seems to be in the movie, will be the reality in well less than 50 years.

Taking 30 steps in a linear fashion will take you from your house to the street.
Taking 30 exponential steps will take you around the world.

30 years ago, the best computer took up a large room, demanded a lot of electricity, cooling, had limited computing capabilities and costs hundreds of millions of dollars.

Today your smart phone is thousands of times smaller, costs a few hundred dollars, has thousands of times the computing speed plus hundreds of other uses like INTERNET connection ( huge improvement) a camera, recorder, date book. go-everywhere phone, etc etc etc.

In the exponentially improving AI and AI-tech fields , the jump to atomic or molecular size level computers/ nanobots will not take the 30 years it did to go from room-size to pocket size computers .

It will be something closer to 14 years.

There are countless examples of how very advanced AI and AI-driven technologies can and will build upon each other and this is happening, more or less, right under our noses.

In 18 months there will be a 20 billion neuron equivalent wafer;
in 3 years 40 billion;
in 4.5 years 80 billion;
in 6 years 160 billion.

The human brain has 100 billion neurons. So after four doublings, these wafers will surpass our abilities.

The human brain is immensely complicated (we all know that by now), and about a third of it (possibly more) is involved in processing visual information. Realizing hardware/software that can emulate that level of sophistication is going to take many decades, IF moore’s law does collapse in the next ten years. If it doesn’t, then I guess I could be wrong.

Google recently showed just what the field is up against when they used something like 16,000 processors to do deep, unsupervised learning on massive amounts of visual data. After a few days of processing, they advanced the field by quite a bit, but we’re still talking about 2d, static photos (i believe they were frames taken from youtube videos). Robots in the field would still need to be able to handle moving objects, shadows, partially obscured objects, etc.

The machine capable of doing this will be enormous, given current technology. Imagine trying to fit that into a robotic body, or imagine supercomputers trying to remotely feed commands to many robots.

I suppose that if neuromorphic technologies work out as it’s hoped, then moore’s law won’t even be an issue, and we might have truly autonomous machines within a couple decades; however, that type of hardware is proving to be extremely difficult to construct to be reliable and deterministic, and I’d say it’s still a wildcard.

http://www.youtube.com/watch?v=N_m56irWKeI&feature=related

check out this robot then… http://www.youtube.com/watch?v=VsnXU_JsnsQ&feature=watch_response

why not merge this with a real person?

like a version of this http://www.kurzweilai.net/double-by-double-robotics

As sad as it may seem, in my dotage, I think I’d trust a robot more than a human to care for me.