Subject:
|
Re: Design
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Sat, 3 Dec 2005 21:31:15 GMT
|
Original-From:
|
dan miller <danbmil99@yahooSPAMLESS.com>
|
Viewed:
|
1464 times
|
| |
|
|
--- PeterBalch <PeterBalch@compuserve.com> wrote:
> I've been
> designing autonous robots for the retail market and I've never ever made a
> profit. (Even though some are in production.) Every other designer I've
> met
> has admitted the same result. Why does Robosapiens have a remote-control?
I can think of one field where autonomous robot-like, pseudo-AI is alive &
kicking, and making tons of money. One word: videogames.
I've been fascinated with the relationship between simulation environments
(ie videogame-like scenarios) and the real world. I've got lots to say
about this, but to the point at hand, people will accept autonomous robots
when they do something people care about. But for the forseeable future,
robots will be slaves to human beings (as animals were, and unfortunately as
some people were to others). We will need to interact to tell the robot
what we want it to do as our servant. While remote controls will stay
around (IMHO -- at least until voice rec is really mature), the commands to
give the robot will become progressively more involved, and require a
greater amount of autonomy.
Think of how you control a character in a typical 3D RPG game. You push the
joystick forward, and the character starts walking in the direction you
indicate. You click on an object, and the character picks it up. These are
very sophisticated actions, which real-life robotics is no where near
achieving (outside of some carefully crafted lab situations).
In contrast, the autonomous characters in these games are truly autonomous;
the controlled and perfectly known state of the virtual world allows these
agents to act within their world in very sophisticated ways. But the point
to remember is, they have goals that drive their behavior -- whatever the
game scenario implies. The Wizard imparts wisdom; the warrior fights, and
so on. They are given virtual personalities by the programmer. To the
gamer (if it's well done), they do indeed seem like living creatures, with
ideas in their heads. That is a projection, but it's a reasonable one to
make given that we don't have direct access to their source code.
IMSHO, the problem with real-world robotics is simple in principle (but
difficult in practice): we have to be able to learn enough about the real
world to act reliably within it. Then it will be an easy matter to program
up a 'bot to do all sorts of useful things -- get a beer, mow the lawn, etc.
The Roomba is an excellent first step, because they constrained the problem
to the point where they can accomplish the task well enough for a human user
to be satisfied.
I believe that retail robots can only succeed when they solve real-world
problems reliably. In most cases, that's a very tall order. But I think
it's unfair to say the robotics & AI community are asleep at the wheel.
Until recently, I was doing research at CMU; initially in CS, but I ended up
getting involved in a robotics project. Though he wasn't involved in this
project, I met Randy Sargent, who invented the programmable brick (ie RCX)
as a thesis at MIT. Trust me, this guy is no dope; and his work with
BotBall (KISS) is really quite amazing. Some of my students had done
Botball, and it was a great grounding in putting theory to practice. Yes,
it's a toy world, but it's a more sophisticated world by far than the world
robots were able to navigate 20 years ago.
As for the Darpa challenge (as you all probably know, we lost by a hair to
Stanford), I find the reaction very interesting. Recently a friend of mine
who was in the pits during the race gave a talk, and the general take
(including his) was that somehow everybody wimped out. The GPS was too
accurate; everyone used the same lidar (SICK); DARPA dumbed down the race to
make sure there were winners, etc. To me, it sounded exactly like the
griping I've heard about Deep Blue (another CMU venture originally) -- that
it won the contest in a brain-dead way; there were no conceptual
breakthroughs; etc.
While there may be grains of truth in both cases, it sometimes seems like as
soon as machines achieve something, we just move the bar up and say "well,
that one was easy -- but can it do this?" I honestly feel that the deep
issue is a discomfort with the idea that many of the tasks that make us feel
deeply human turn out to be not so special after all. We want to believe
there is magic in the broth; but maybe it's just stone soup -- the magic
isn't in any one ingredient; it's in the mix.
my rant for today -dbm
__________________________________________
Yahoo! DSL – Something to write home about.
Just $16.99/mo. or less.
dsl.yahoo.com
|
|
Message is in Reply To:
 | | Re: Design
|
| Bruce (...) Yes that's true and very sad. It's a problem I'm interested in because for a few years now, I've been designing autonous robots for the retail market and I've never ever made a profit. (Even though some are in production.) Every other (...) (19 years ago, 3-Dec-05, to lugnet.robotics)
|
3 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
