From: Daniel Burfoot (daniel.burfoot@gmail.com)
Date: Fri Aug 17 2007 - 04:05:37 MDT
I'd like to comment briefly on a topic that was touched in the previous
thread: the relationship between robotics and AI. I'm an AI researcher in a
robotics lab, so I spend some time thinking about this issue.
As noted, it's not immediately clear why AI researchers should be interested
in robotics. If we say that our goal is "human-level intelligence", we can
say that this goal involves many subgoals (X1, X2, X3.... XN) where the
various Xi stand for things like:
1) Vision
2) Language
3) Planning
4) Mathematics
5) Morality
...
And of course each of these is individually difficult and may involve
several of its own subsubgoals (or subsubproblems). Now, we would hope that
achieving one of the subgoals will greatly expedite the discovery of the
others. For example, many people think that vision and language have some
similarities: both involve "parsing" of ambiguous data. If that is the case,
then a smart ordering of the subgoals could significantly speed up the
realization of the entire set. Unfortunately it is not at all clear how to
do this.
The view of those who think robotics is important for AI is that researchers
should basically follow the same path as evolution in achieving the
subgoals. That is to say, we should first build machines that can
pseudo-navigate their surroundings, make basic sensor readings and respond
to them, etc. Once that is achieved, we can hopefully move on to more
complex tasks, building up to a human-like machine that can walk around,
perform basic repetitive chores, etc. Then potentially we can make the leap
to real human thought and culture.
This view of how to order the subgoals has been advocated most strongly by
Rodney Brooks (google scholar for "Intelligence without Reason" and
"Intelligence without Representation"). The Roomba is a direct descendant of
this line of thought.
Now the question is: does anyone have a better way to order the subgoals?
An alternate rationale for pursuing robotics is that there may be some
practical applications along the way. I wouldn't mind having a robotic
chaffeur, for example. Some people might dismiss these quotidian
applications as insignificant compared to the Singularity, and I would tend
to agree. Note however that video games are insignificant compared to
computing, nevertheless video games played an important role in driving the
development of computers.
Dan
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:58 MDT