Interesting Register Article - Will Robots Ever Be Like Humans?
The Register is an internet publication I regularly read which discusses technology and other areas of interest from an editorial point of view. It is not always the most accurate and impartial place to read things, but it is often quite fun and interesting. I have been on a reading spree and come across a number of articles that I think are somewhat relevant in robotics.
Some of the most fun ones are those based upon reader questions, which can sometimes live up to the adage of “ask a silly question, get a silly answer”, but can sometimes be insightful and informative. These have all been stumbled upon.
Will Robots Ever Be Like Humans?
The article is interesting because it actually cites a number of sources. The main gist of the answer is not so much about the AI development but more the actual physical body structure. I would suspect that having human like dexterity and suppleness may actually be closer than having human like intelligence.
Effectively and adaptively controlling so many actuators may require a vaster intelligence and servo motor control than artificial intelligence currently offers, however, much like the Honda Asimo in the current adverts, it could be preprogrammed with a few very lifelike activities.
Is an artificial eye close to reality?
This is actually an article covering a number of small questions with short answers and citations to back them up. The actual concept of an artificial eye is not new, and students have been known to perform activities using electrodes and grids of light sensors.
While these basics have shown some use, and have even been used by blindfolded students to navigate a maze, more advanced CCD’s (the kind of technology used in every mobile phone camera and webcam) have become very common and cheap. With some additional processing, they could be used. However, I also think this arrangement is not enough to constitute an “eye”. An adjustable focus wide angle lens, with varying detail bands, and a system to allow it to be moved (not fixed position) would make a simple CCD more eye-like.
Can colours be detected by touch
Another question answered in this article is if colours can be detected by touch. Now while that seems not only unlikely, but probably impossible, in certain light conditions, the reflected vs absorbed heat may be a rough indicator of the shade of the colour on the item. Most people have experienced touching a matt black item that was that much hotter than its surroundings on a hot summer day. Unlikely, but also more plausible than any “psychic” explanations.
The Register have cited a reference to a lady who believes she can do this. Of course, it is very likely that most people who claim this are little more than elaborate (and not always so elaborate) hoaxes.
I don’t know if I’ve mentioned this site before, but I stumbled on it years a go, and stumbled upon it again today when looking for a herb identification key on line (which I did not really find). 20 questions is a database based upon question-answer responses, and it can identify what you are thinking about, a variation on the animal, mineral, vegetable game if you like.
It is actually fairly astute, and can guess most things within 20 questions. It is a little fickle, as it learns from its players, you may find your answers contradict those of “most players” - which do not necessarily mean you are wrong, but in a way, learning to identify it as most players see it is probably more valuable. They also sell a 20q gadget which can be used to impress people with a little bit of AI trickery.
It is also a bit of good harmless fun.
This is a rather odd and amazing concept. It is the idea of offering an API that allows you to set tasks that an AI is not yet capable of, like identifying objects in photographs, and for those who perform such tasks, an opportunity for a quick mini-earner. It is not a lot, normally measured in cents, and for non-US people it earns Amazon vouchers.
The system is currently a Beta, run by Amazon. I have played with it, and think it is a concept that could be very handy for tagging and creating a lot of meta information, thus creating the “semantic web”. Link concepts like this with the 20q engine - and you have a powerful expert system involving people as well as machines.
It is actually named after an infamous 18th century chess machine, which appeared to play a really good game, but turned out to be a man curled up, a great hoax. This was known as the Turk, and by facilitating tasks that people would like a computer to be able to do, MTurk creates an automated way of linking up capable humans with tasks required by others or even by machines.