Accomplishing DARPA's robot challenge (pt 3) using
Human Level Artificial Intelligence
Note: To make this website free to the public
please click on an ad to support my sponsors or you
can make a tax-deductable donation using Paypal
(click on the donation icon on the left).
This video shows a
robot making a toy model. I know it's kind of
childish to make toy models, but the point of this
video is to show a robot using various tools to make
things. The robot is using various tools: razors,
glue, styrofoam, q-tip, a sticker, an electric
cutter, and paint brush to make things. There are no
sound in the video because i wanted to show the
viewers what the robot is thinking while crafting a
toy model. The flashing text and freeze frames are
the internal thoughts of the robot and are not
instruction text for the viewers.
The video shows the
hierarchical steps the robot has to go through in
order to modify a toy model. There are 2 tasks the
robot has to do: 1. turn the toy model into a damaged
toy model. 2. make the dirt environment model. Within
each task, there are more steps involved. The robot
has to use his mind to do tasks recursively and
linearly.
The big question is
how exactly does the robot know how to use different
tools???? All knowledge from the robot's brain comes
from teachers in school or through reading books. No
programmers are needed to manually insert knowledge
into the robot's brain. No machine learning is
required to learn things automatically.
The key to this type
of learning is: the robot's brain has to store
information from the environment through its' 5
senses, frame-by-frame. When the robot walks, he is
storing frame-by-frame body movements, visual data,
sound data, smell data, etc. When the robot is
balancing on a bike, he is storing frame-by-frame
body movements, visual data, sound data, etc. I call
these movie sequences, pathways. The robot's brain is
storing movie pathways in memory. In fact, the whole
idea behind a human robot, is to build a brain that
can store, modify, organize, and delete movie clips.
The brain has to store 50-100 years of movie clips --
every second the robot exists his brain is storing
information on the environment frame-by-frame.
The key is the
robot's brain forget information. This is how it can
store 50-100 years worth of continuous movie clips.
Information in stored movie sequences degrade -- its
quality turns fuzzy. In some cases, no movie sequence
is stored in memory, only words and sentences to
represent an event. Other movie clips in memory are
so degraded that only a few pixels represent an
event.
The robot's brain
keeps good pathways and forgets bad pathways. For
example, when the robot walks, only the good pathways
to balance and walk are kept. The pathways that lead
to falling down or walking abnormally are forgotten
from memory. Another example is riding a bike. The
pathways that allow the robot to ride a bike are kept
in memory. These pathways include balancing the bike
and controlled navigation. The bad pathways are
knowledge that leads to falling off a bike or
uncontrolled navigation. In current robots (2013)
programmers insert algorithms and complex math
equations to balance a robot or ride a bike. My human
robot doesn't require math equations to balance while
walking. My robot is taught how to balance and walk
through guidance and logic and the good ways of
walking are stored as pathways in memory and the bad
ways of walking are forgotten from memory (pain and
pleasure plays a role in this type of learning).
Understanding
physical objects in 3-d space and how objects
interact with each other is based on experiencing
them, frame-by-frame. Humans understand 3-d because
we see things frame-by-frame. We know how a cat looks
from the front, side, and back because of movie
sequences of the cat from different angles and view
points. We know how a pen feels when we grab it. We
know how much pressure to put on the pen to write
letters on paper. We know how to use a electric saw
because we experienced it through our 5 senses,
frame-by-frame.
This is important
because the robot is able to use any tool (billions
of different tools) to do tasks. The robot learned
how to use pens, scissors, razors, hammer, glue, a
paint brush, etc from teachers in school. The robot's
brain has stored pathways that will tell the robot:
how does the tool feel, what are the consequences of
a tool's actions, how much pressure should i put on a
tool, how do i balance the tool in my hand, how do i
control a tool to make a cut, how do i take a sticker
off, how much pressure do i put on sticker, etc. The
pathways in the robot's brain stores all the
information about how to use a tool, frame-by-frame.
The good ways and controlled ways of using a tool are
stored in memory. The bad ways of controlling a tool
are forgotten from memory.
Also, the pathways
are structured hierarchically, whereby similar
hierarchically pathways are fused together (or
shared). This prevents repeated information from
being stored in memory.