ROBOT GETS SMARTER BY ASKING FOR HELP (published by NewScientist).

“Object recognition has long troubled AI researchers. While computers can be taught to recognise simple objects, such as pens or mugs, they often make mistakes when the lighting conditions or viewing angle change. This makes it difficult to create robots that can navigate safely around buildings and interact with objects, a problem Willow Garage encountered when building its Personal Robot 2 (PR2).

Where AI struggles, humans excel, finding this sort of recognition task almost effortless. So Alex Sorokin, a computer scientist at the University of Illinois at Urbana-Champaign, who collaborates with Willow Garage, decided to take advantage of this by building a system that allows PR2 to ask humans for help.

The system uses Amazon's Mechanical Turk, an online marketplace which pairs up workers with employers that have simple tasks they need completing. The robot takes a photo of the object it doesn't recognise and sends it to Mechanical Turk. Workers can then use Sorokin's software to draw an outline around an object in the image and attach a name to it, getting paid between 3 and 15 cents for each image they process.

In initial tests, the robot moved through Willow Garage's offices, sending images to be processed every few seconds. Labelled images started coming back a few minutes later. The accuracy rate was only 80 per cent, but Sorokin says this can be improved by paying other workers to verify that the responses are valid.

Sorokin believes his system will help robots learn about new environments. A cleaning robot, for example, could spend its first week in a new building taking pictures and having people label them, helping it to build up a model of the space and the objects it contained. If it got stuck, it could always ask for help again.

“This is a fantastic idea,” says John Leonard, a roboticist at the Massachusetts Institute of Technology. Potentially this could allow robots to operate for long periods without direct intervention from a human operator, he adds.”

ROBOT GETS SMARTER BY ASKING FOR HELP (published by NewScientist).

“Object recognition has long troubled AI researchers. While computers can be taught to recognise simple objects, such as pens or mugs, they often make mistakes when the lighting conditions or viewing angle change. This makes it difficult to create robots that can navigate safely around buildings and interact with objects, a problem Willow Garage encountered when building its Personal Robot 2 (PR2).

Where AI struggles, humans excel, finding this sort of recognition task almost effortless. So Alex Sorokin, a computer scientist at the University of Illinois at Urbana-Champaign, who collaborates with Willow Garage, decided to take advantage of this by building a system that allows PR2 to ask humans for help.

The system uses Amazon's Mechanical Turk, an online marketplace which pairs up workers with employers that have simple tasks they need completing. The robot takes a photo of the object it doesn't recognise and sends it to Mechanical Turk. Workers can then use Sorokin's software to draw an outline around an object in the image and attach a name to it, getting paid between 3 and 15 cents for each image they process.

In initial tests, the robot moved through Willow Garage's offices, sending images to be processed every few seconds. Labelled images started coming back a few minutes later. The accuracy rate was only 80 per cent, but Sorokin says this can be improved by paying other workers to verify that the responses are valid.

Sorokin believes his system will help robots learn about new environments. A cleaning robot, for example, could spend its first week in a new building taking pictures and having people label them, helping it to build up a model of the space and the objects it contained. If it got stuck, it could always ask for help again.

“This is a fantastic idea,” says John Leonard, a roboticist at the Massachusetts Institute of Technology. Potentially this could allow robots to operate for long periods without direct intervention from a human operator, he adds.”

http://transumanar.com/index.php/site/next_cosmic_engineers_meeting_in_second_life_discussion_of_ctrl_alt_r_rebak/

“Generally speaking, we only notice technology when it fails us.”

“Looking at the first trend, there’s little doubt that we are rapidly approaching an era where the scale of information technology grows beyond a human’s capacity to comprehend.”

“Computers, though, are still not quite as ‘invisible’ as the TV set is, and that’s because they are not yet ‘plug and play’. I think most people switch on the computer, half expecting it to not boot, fail to connect to the Internet, drop their work down a metaphorical black hole and so on.”

These all resonate with me, in terms of my technology experiences! It is hard to be immersed when your computer crashes repeatedly in the middle of a SL session!

Great essay, and some wonderful replies. I really enjoyed reading this. I need more of this, and less time reading People magazine on line! (at least I have made the break from television. We don’t have cable. Haven’t had it since late 2005. We only use the tv to play video games, and sometimes watch a movie.)

Princess Ivory

At the risk of undercutting my own argument somewhat………..I found this on Google.

“Is Google making us Stupid.”

http://www.theatlantic.com/doc/200807/google

Google is great for finding the answers to questions like ‘when was the first chess playing computer invented’. Actually, I do not know so let us ask it…

Ok, so the first analogue chess computer was invented in 1911 (or 1912, historians are not sure), and the first chess playing program to run on computers as we know them was invented in 1951 by Dr. Dietrich Prinz.

Now, I could commit that information to memory, but why should I bother? After all, Google is so much more efficient at recalling such facts; I might as well just ask it to remember for me whenever I need to know.

What if Google was with me wherever I went, whispering exposition in my ear, displaying answers in my visual field? From all outward appearances, I would have expansive knowledge, but do I really KNOW anything? This is not a new anxiety, because Socrates said of the then quite recent invention of the alphabet,

‘(It) will create forgetfulness in the learners’ souls, because they will not use their memories; they will trust to the external written characters and not remember of themselves. The specific which you have discovered is an aid not to memory, but to reminiscence, and you give your disciples not truth, but only the semblance of truth; they will be hearers of many things and will have learned nothing; they will appear to be omniscient and will generally know nothing; they will be tiresome company, having the show of wisdom without the reality’.

(Yes, I asked Google to remember that quote for me).

Obviously, for a lot of pattern recognition tasks you are much better off using your own brain because AI is just so profoundly inferior. No way would I ask an AI to identify objects like ‘tables’ or ‘chairs’, my brain does a great job, thanks. And while Google translator is OK, it is clearly still advantageous to learn foreign languages.

But, what if those pattern-recognition capabilities are aquired by search engines, and then coupled with their ability to remember at electronic speeds? What if Google was BETTER at putting names to faces, translating languages, recalling the last place I left my keys…Would I then come to rely on it to do ALL my recollection for me? Would a large portion of my mind exist in the Cloud, leaving my old meatbrain largely empty and unused?

“extract the salient details”

‘These are indeed the pressing issues’.

Right. Sounds easy, but in practice it is very difficult indeed. Whenever you hear somebody claim AI will be on a par with humans within a few decades, you can be sure that person is not a neuroscientist (s/he is probably a roboticist or something like that). While some neuroscientists freely admit the brain will be reverse engineered into software some day, I know of no neuroscientist who expects that to happen in under a generation.

‘I do not mean to be overly querulous’.

Aww! I like your comments:)

Well…..um…….. ok.
In which case bring on the implants, because I swear, the more I use google………the stupider I get.

” I really don’t see why there wouldn’t be progress towards currently unsolved goals that nature has shown are solvable in principle.”

Agreed. !!

“More data will not lead to intelligence, but what about more of the right kind of data? ”

“extract the salient details”

These are indeed the pressing issues.
I merely note in passing, that these also involve subjective as well as objective judgements.

I do not mean to be overly querulous, in fact your essays have always offered much food for thought.

More data would not lead to intelligence. This is almost like the claim that computers will become intelligent if only we can increase their raw power; that Moore’s Law progresses until…!!Bingo!!..computers are intelligent. Although Ray Kurzweil is often portrayed as believing this by AI skepics, it is not actually his position. He has stated several times that the continuation of Moore’s Law is a necessary but not sufficient condition for creating artificial general intelligence.

More data will not lead to intelligence, but what about more of the right kind of data? What about improvements in our ability to study living brains and extract the salient details as they process information? According to Kurzweil, ‘The pace of working models and simulations is only slightly behind the availability of brain-scanning and neuron structure information’.There are over 50,000 neuroscientists in the world writing articles for more than three hundred journals, as well as scientists and engineers involved in the development of new and improved scanning and sensing technologies, and the likes of Google constantly seeking ways of finding high level knowledge amongst low level information.

OK, so once again I have gone and used the word ‘improvement’ a lot. Admittedly, it is probably true that a lot of theories regarding the emergence of intelligence are innacurate. But the fact that there are so many different approaches to artificial intelligence can be seen as positive from an evolutionary perspective. After all, nature throws up many possible solutions to any given problem and then weeds out the less effective approaches.

If information technology can be developed so that humans and computers become better at searching probability space for effective solutions to any given problem, I really don’t see why there wouldn’t be progress towards currently unsolved goals that nature has shown are solvable in principle.

Your essays often use a similar form.

1. Where we are at now.

2. Use of the word “increasingly” or “progressively”

3. Bingo !! The Singularity !!!

Of course, this may well be true, although I remain to be convinced that more data leads to more intelligence.

It kinda reminds me of the difference between Scales and Music.

Music is made up of scales, each with their own laws and states of tension and release with each other. They are the data building blocks of music.( along with sundry others, Rhythm, Harmony, Timbre etc.)

Ok, lets see if this works.

1. Millions of Pianists across the world are practising scales.

2. Due to technological progress, they “increasingly” improve their technique. They “Progressively get faster and smoother at their execution of playing scales.

3. Bingo !!. Rachmaninoff’s 2nd Piano Concerto !!

Of course Sergei managed to do exactly this, without the aid of Google.

The practising of Scales, may indeed have helped him.
But it was not the Prime Mover, in terms of the unleashing of his creativity.

The most persistent problem at the moment, is seperating the Signal from the Noise,and here we enter a largely subjective realm, where my signal may be your noise.

What our hypothetical AI considers to be Signal rather than Noise, remains the fascinating if unknowable question.