Ray Kurzweil Has Some Incredible Predictions for the years up to 2045

[accelerator]

Hi Guys

Ray Kurzweil is a US inventor and Artificial Intelligence expert. Bill Gates calls him “the best person I know at predicting the future of artificial intelligence.” He has a good track record at predicting the arrival of future technologies, and this article gives some of his incredible predictions for the near term future up to 2045. We live in interesting times.

http://singularityhub.com/2015/01/26/ray-kurzweils-mind-boggling-predictions-for-the-next-25-years/

 

[Edwin]

His book "The Age of Spiritual Machines" is a good read.

 

[humble pie]

Just touching on one of his predictions in AI you may find insightful..

he predicted…that a computer would defeat a world chess champion by 1998. Then in 1997, IBM’s Deep Blue defeated Garry Kasparov

As someone with an above average nerdy insight into Chess & AI (I work with both nearly every day) here is a quick heads up, so you don't have to worry about Schwarzenegger/Terminator type future yet as AI progress (and technology progress in general) is still lame really.

Chess is considered one of, if not the purest form of measuring computer AI progress. The number of Chess moves available in a single game of chess is 10 to the power of 120 - to give you some idea of the size of this number, it's larger than the number of seconds in the known universe since time began, so chess is not solvable by a machine (unlike Draughts which has been solved).

A chess computer wins games by brute-force calculation as it cannot see the board, it has no *strategy* unlike humans who can see the board and have a 'plan' as well as calculation ability. (I don't want to go to deep but have a look at Moravec's paradox if you are really bored).

The point is, in the wider sense, as long as a human being can win or draw just one game against a chess computer (at any given point in time) then AI is not dominating human race

Garry Kasparov did beat Deep blue during in that match and there has not been a Chess computer that a human being has not been able to beat or draw against yet.

Hope that made some sense

 

[RobM]

I've been interested in Kurzweil and the singularity for a long long time. Luckily it won't matter if he's right (Personally I think he is).

 

[Edwin]

(I don't want to go to deep but have a look at Moravec's paradox if you are really bored).

Although I'd not heard of it until you mentioned it, there was a great example of Moravec's paradox in the Darpa Robotics Challenge http://www.theroboticschallenge.org/ which was held over Friday/Saturday this week. In it, robots had to drive a (modified) jeep, get out, open a door, turn a valve, pick up a power saw and cut a hole in plasterboard (around a guide), unplug a cable and plug it into a different socket, clamber over (or shove through) some "debris" and climb 5 stairs.

The whole event was streamed live, and it was fascinating viewing, even though a lot of the time there was so little action that it was like watching paint dry. Three robots took home prizes, and all 3 managed to complete all 8 tasks within an hour (there were another 21 teams that achieved 7 or fewer tasks). I'm guessing any average human could have finished the course in 2 minutes max - and without needing a team of engineers to send commands, pick them up when they fell over https://www.youtube.com/watch?v=7A_QPGcjrh0 (10 minute penalty applied) etc. etc.

Over time the robots are only going to get better and better, but because the real world's full of awkward things like gravity, wind, rain, sharp angles, unstable surfaces, etc. etc. it's probably something where improvement will be relatively slow.

On the other hand, the Darpa self-driving car challenge http://en.wikipedia.org/wiki/DARPA_Grand_Challenge started off amazingly badly in 2004 but then technology advanced very quickly in subsequent years of the competition, and led to spin-offs like Google's autonomous vehicles which have driven over 1,000,000 miles without any computer-caused accidents.

What's the difference? Well, trying to get a bi-pedal(ish) robot to walk through a simulated urban environment is incredibly hard because it requires the equivalent of fine motor skills - but getting a car to drive itself safely is very much a brute-force computation exercise. The car's in no danger of falling over, so the actual physics behind its movement are very very simple compared to trying to keep a robot balanced as it walks and navigates obstacles, climbs stairs etc. The "big ask" is the ability to see/interpret the road ahead, and to analyse and respond to hazards/dangers/situations - all things that improve very quickly as you throw more computing power at the problem.