Now try to imagine what it would be like to have a Weed-bot roaming about with the ability to size up objects and distances better than any human and to give a list of all objects in the surroundings better than any human (whether it can name them or not). Obviously such a cart would be invaluable in exploration on other planets where objects not found on Earth may exist. On Planet X one needs the ability to isolate objects from their surroundings (ie figure-ground ability) and the ability to measure their size and shape and distances from their surroundings.
Next imagine such a robot with programs for all of the arithmetic-logic-mathematics on Earth. It can then perform all of the logical manipulations which humankind collectively could perform on Planet X objects. For example, it could generate a cross-section of a new geological formation in any plane limited by the topographical and sub-surface meaures taken to date. Since logic has its own language, it could communicate all of that information to humankind. ... in a way which surpasses human intelligence. So this is our second general SHAI. Like OCW-bot it could be created now. We have the knowledge and technology to do so. "All" we need is a few billion dollars ... the price of a modern aircraft carrier. And perhaps DARPA (Defence Advanced Research Projects Agency) has one but is not telling the public.
<http://www.britannica.com>
Encyclopedia Britannica gives an overview of human sensory reception The prospect of what Kurzweil calls a general "virtual sensorium" (1999, p.311) is at hand. The "Perception Online Journal" provides more technical
information on development in this important field:
And vision is the 'king of the senses' when it comes to the prospect for developing a virtual sensorium which enables super-human AI. Some aspects of machine vision are obviously beyond the capabilities of humankind now. Machine eyes have tbe power of microscopes and telescopes, which human eyes do not have. Professor Arun Shukla of the University of Rhode Island reported on Discovery television, Sept. 13/00 that digital cameras can now "image" a scene millions of times faster than a human. E.R. Davies writes in his 1997 text on machine vision that "...there seems to be no reason to suppose that the eye is the ultimate vision machine ... clearly it is worth being aware of the possibility of the robot eye having capabilities superior to those of biological systems." (p. 14).
GIVING ROBOTS THE GIFT OF SIGHT
Davies adds the following: "...there is an important guiding principle: if the eye can do it, so can the machine...." (p. 14). For more information about machine vision online refer to <http://www.qub.ac.uk/ivs>
and Hans Moravec at CMU is working on a 3-D vision system that "...will allow machines to make their way through offices and homes" according to an article by Ed Frauenheim, CNET News.com, Dec. 30/02, titled "Giving Robots the Gift of Sight". <http://c.moreover.com/click/here.pl?e55729046&e=6507>
But "HelpMate", the courier robot developed by Engelberger (dubbed the father of modern robotics with Devol) now roams through the corridors and elevators of over 100 hospitals worldwide using range finding ability to avoid collisions and Sony's 23 inch SDR-4X bipedal humanoid can recognize human faces.
The great challenge is to make the range finding of a Helpmate so compact and extensive that it can measure up any and all objects and distances; and to make the object recognition of an SDR-4X so powerful that it can recognize all objects and not just faces.
There is a challenge we could put forward to Omron, a private company which presents its "Solve It" challenge to the public. They say, "Give us your toughest application or components problem and we'll solve it for you." Here we have the problem, and the solution is certainly worthy of a Nobel Prize: DEVELOP A COMPACT MACHINE VISION SYSTEM SUITABLE FOR A BIPEDAL HUMANOID ROBOT WHICH SURPASSES HUMAN EQUIVALENCY IN SIZE-DISTANCE DETECTION AND GENERAL OBJECT RECOGNITION. <http://www.solveit.omron.com>
ROBOTIC SHOW AND TELL
And of the two components to that challenge above, general object recognition is the more challenging.
A widely used psychological test called the Peabody Picture Vocabulary test has one of the greatest age-spans of any measure for human intelligence. Norms are available for ages 2-90+, yielding an IQ-like mean of 100 with a standard deviation of 15. Remember that human verbal ability usually begins in year 1-2 and at first it is only babbling. See <http://www.psycan.com>