Artificial
Intelligence... A theory but no answer
Over the past couple of decades much work, and progress, has been
made in the pursuit of creating artificial minds and lifelike humanoid
movement. Researchers from many different nations have experimented
with recreating realistic human characteristics - with some success.
Honda's
humanoid robot ASIMO leads the way in mobility, the 120 cm high
robot has already mastered the complicated balancing act required
when walking, running and jumping and it can safely navigate around
random objects. ASIMO is also capable of recognising people and faces
and greets people accordingly, he can also move where indicated and
recognise his name. A remarkable machine no doubt, but ASIMO is not
truly intelligent.
Work has also been done on facial recognition robots, these complex
cameras are capable of recognising people and emotions and copying
or reacting to facial stimuli. But, again these robots are merely
fullfilling their programmers desires. However the technology being
developed here is important to future more complicated robots, these
robots will be capable of decision making, reasoning and spontaneity
and even genuine humor amongst other 'human' traits.
Artificially intelligent robots in my personal and unqualified opinion,
and it is just that, an opinion, should be programmed by trial and
error and by example, exactly as humans and animals learn. The only
things that are man made are the separate components.
The eyes for example, complicated recognition cameras, already in
development, able to recognise color, depth, movement, speed, and
light and then by using the technology being developed with the facial
recognition robots the central processor is capable of sorting images
into order of priority for its given task or situation the priority
will be changing from second to second.
Imagine walking down a busy street your eyes are constantly scanning
and checking your path is clear, while also trying to determine which
store you are going to go into next, and you are also looking out
for someone you know. Each second has its own priority do you side
step to the left or right? sometimes you get it wrong, or others,
and you bump into people, errors are a part of human life. Why not
use the same principle with robots?
Example Scenario
The mobility and human appearance of the robot in question need not
be great, however that is the simplest part, and will likely be far
more advanced than the first artificial brain. The learning process
should be just like that of a child, someone spends time every day
for hours teaching and showing the robot various techniques for life
starting with simple colours, shapes and movements like a baby, then
moving on to simple interactions and tasks like an infant, before
gradually increasing the complexity of the tasks while integrating
them into a series of tasks and completing complex maneuvers.
For this example lets say the robot can walk outside easily, he has
been shown and guided around the streets for several years and has
many experiences and problem solving solutions stored in his memory.
He has been shown a chocolate bar and he has been told to go out and
buy one from a nearby newsagents. The robot has been shown the outside
many times before and he can recognise sidewalks and knows this is
safe, however over the years he has also been shown that it is safe
to walk on the road to cross if there is no oncoming cars or the traffic
is stopped (as well as many other countless variables that can only
be learnt with time - growing up, for humans). Lets say he is walking
along to buy this chocolate bar and there is road works ahead, he
has never seen these particular roadworks before and they are blocking
his path, the newsagent is on the same side of the road as him, his
options are:
1. He must either walk on the road until he is past the roadworks
and then get back on the sidewalk.
2. He can cross the road continue on past the road works and then
cross back over to the side with the newsagents.
3. He can turn around and go back, because he is not sure if his trainers
(parents) want him walking on the road on his own.
Everyone has made decisions similar to these but we weigh up the pros
and cons of each option and hope that the one we choose is the right
one. In this case the robot decides that there hasn't been any cars
that have gone past him in a long time, something he had remembered
in his short term memory bank. So he checks the road and sees there
is actually a car approaching he judges its speed by comparing its
size to a couple of video frames previously and waits for it to pass.
Unfazed, it is only one car, he steps out onto the road, he knows
this is a dangerous place but all safety precautions have been taken.
He continues along the road staying close to the roadworks, he has
seen that is what other people do in similar situations then after
a few metres he is past the roadworks and he steps back up onto the
sidewalk and continues on to the shop.
On the way he notices that there is a new, closer newsagents which
has recently opened since he was last here. He recognised that it
was a newsagents because of the A-board sign out in front and various
signs on the windows also the small cluttered size of the premises,
all clues to a newsagent, and he has recognised various stimuli generally
associated with newsagents.
After entering he searches for the chocolate bar, he can remember
what that looks like because he was shown it just before and has retained
the information for the time being. He finds the chocolate bar and
takes it to the counter, he knows this to be the counter because there
is a till shaped object with a man facing him. He buys the chocolate
bar and walks back the way he came.
Conclusion
Because this was one of his first forays into the world on his own
it was a simple task, however many thousands, even millions, of tiny
calculations took place. The importance of road safety has been drilled
into him for years on end, and this was his most important decision
to make. Although he took a more risky course of action it was by
all factors quite safe overall, something he has learnt with time
and experience.
The recognition factors mentioned are not infallible, however everybody
has miscalculated and misjudged at some point in their lives, usually
many points. Everybody has walked up to a fellow customer in a store
mistaking them for a store worker because their clothes are the same
or even there stance is one of a salesperson, not browsing customer.
So a mistake here and there is not failure, it just takes time, experience
and trial and error to see as many scenarios as possible and remember
the successful solution, reviewing similar experiences to find answers
to new problems.
Now just the simple matter of building it.
|
