You can use the Spider
http://www.lantronix.com/it-management/kvm-over-ip/securelinx-spider.html

If you carry an AP on your bot you can use it via wireless.

Alberto

On Wed, 2009-02-25 at 18:23 +0100, Giuseppe Marullo wrote:
> Jim,
> I have one beautiful 1U DSR2030 ip kvm with a huge intelligent power
> socket unit (IPDU something, 20 sockets, friendly named Rocco
> Siffredi,
> 2m long), but it is far from portable.
>
> Newerthless, I could be interested in a compact wireless portable
> unit.
> A new market opportunity?
>
> Giuseppe
>
> Jim McBride wrote:
> >
> > Greetings,
> > Is anyone aware of a (portable) product that would allow you to use
> > the keyboard and monitor of a laptop with another computer? I am NOT
> > talking about Remote Desktop, Netsupport, Timbuktoo, KVM
> > switch(unless it was portable).... It sure would be great to be able
> > to use a laptop to watch a robot/computer boot up and simply use a
> > laptop computer as the terminal for windowz. I get tired of luggin
> > around a monitor/keyboard/mouse.
> >
> > I am just about ready to just get one of those all in one rack
> > mounted collapsible keyboard/monitor setups and modify it to a
> > portable unit.
> >
> > JIMc
> > x22661
> > National
> > Ignition
> > Facility
> >
> > [Non-text portions of this message have been removed]
> >
> >
>
>
>
>
>

would synergy work for you?

  It works great (server on Linux) but requires both computers to actually run
synergy, so you wouldn't be able to use it for the bootup sequence.

Have fun,
Markus


----- Original Message ----
From: Jim McBride
Greetings,
Is anyone aware of a (portable) product that would allow you to use
the keyboard and monitor of a laptop with another computer? I am NOT
talking about Remote Desktop, Netsupport, Timbuktoo, KVM
switch(unless it was portable).... It sure would be great to be able
to use a laptop to watch a robot/computer boot up and simply use a
laptop computer as the terminal for windowz. I get tired of luggin
around a monitor/keyboard/mouse.

I am just about ready to just get one of those all in one rack
mounted collapsible keyboard/monitor setups and modify it to a portable unit.

JIMc
x22661
National
Ignition
Facility



[Non-text portions of this message have been removed]



------------------------------------

Visit the SRS Website at http://www.seattlerobotics.orgYahoo! Groups Links

Hi Dr. Bruce,

Some thoughts on the topic.

Dave Wyland

--- In SeattleRobotics@yahoogroups.com, "Bruce J Weimer MD"
<bjweimer@...> wrote:
>
> These are the definitions I'm using:
>
> The AIMA Agent [Russell and Norvig 1995, page 33] "An agent is
anything that can be viewed as perceiving its environment through
sensors and acting upon that environment through effectors."
>
> Thus an RC style robot would not be an "agent", but a simple
microcontroller based line follower would qualify.... at least as
an"agent" in these terms.
>
> Proceedings of the Third International Workshop on Agent Theories,
Architectures and Languages 1996 "An autonomous agent is a system
situated within and a part of an environment that senses that
environment and acts on it, over time, in pursuit of its own agenda
and so as to effect what it senses in the future."
>
> Here, however, the microcontroller based line follower fails to meet
the definition because it's doing nothing but following the
instructions in it's program - it's following the programmer's
agenda... not it's own agenda.  Hence it simply isn't autonomous.
>

"And then a miracle occurs." What is "its own agenda?"

1. Where does it come from, if we did not build it in at some level?

2. If we did build it in, is it "its own agenda?"

3. If we set up a method of learning (necessary) and the robot learns
using this agenda, does it have "its own agenda?" Does "its own
agenda" develop over time? If so, when does it emerge? And how do we
tell, objectively?

4. If we do not build it in, where does it come from? Traditionally,
as in Frankenstein and Johnny 5 of "Short Circuit", it is provided
courtesy of a lightening bolt from heaven - a miracle.

IMO, "its own agenda" is in the mind of the beholder and arises
from lack of knowledge that would allow us to predict the robot's
action. Once you can predict a thing's actions, the mystery
disappears. It is then "just a machine."

Try creating  an objective definition of "its own agenda" that
includes the idea of deterministic predictability based on knowledge
of the machine's construction and state at the time of prediction and
excludes adding a random number generator to create mystery.


> In my view, for a robot to be truly autonomous, it might be
programmed with the tools that it needs to do certain things - vision,
memory, emotion, whatever..... but it must not be pre-programmed with
a specific agenda.... it has to develop it's own agenda... just like
each of us did growing up.... and most of us turned out okay!     : - )
>

Problem being that we do not know how we develop our agenda growing
up, psychology and neurophysiology notwithstanding. We have only the
sketchiest philosophical theories on how we develop. They do not
predict behavior very well or consistently.

We have no access to the millions of years of training data, etc. that
form the heredity that guides our development. Likewise, have no
access to the state of the 10^15 synapses in our brain, or to the
similarly astronomical numbers of synapses in simpler animals and
insects.

If we do not know - in sufficient detail - how we got here, our
current internal state and how that state changes with learning, it is
hard to come up with a method for recreating it in a machine.

> Dr. Bruce.
>
>
> [snip]

The laptop can have applications on it. The server, ummm I mean the
ROBOT, cannot.
That spider is interesting too with its browser based interface.


At 10:51 AM 2/25/2009, you wrote:

>would synergy work for you?
>
>It works great (server on Linux) but requires both computers to
>actually run synergy, so you wouldn't be able to use it for the
>bootup sequence.
>
>Have fun,
>Markus
>
>----- Original Message ----
>From: Jim McBride
>Greetings,
>Is anyone aware of a (portable) product that would allow you to use
>the keyboard and monitor of a laptop with another computer? I am NOT
>talking about Remote Desktop, Netsupport, Timbuktoo, KVM
>switch(unless it was portable).... It sure would be great to be able
>to use a laptop to watch a robot/computer boot up and simply use a
>laptop computer as the terminal for windowz. I get tired of luggin
>around a monitor/keyboard/mouse.
>
>I am just about ready to just get one of those all in one rack
>mounted collapsible keyboard/monitor setups and modify it to a portable unit.
>
>JIMc
>x22661
>National
>Ignition
>Facility
>
>[Non-text portions of this message have been removed]
>
>------------------------------------
>
>Visit the SRS Website at <http:// www.
>seattlerobotics.orgYahoo>http:// www. seattlerobotics.orgYahoo! Groups Links
>
>


[Non-text portions of this message have been removed]

--- In SeattleRobotics@yahoogroups.com, "David Wyland" <dcwyland@...>
wrote:
Re: Meaning of autonomous robot

> What is "its own agenda?"
>
> 1. Where does it come from, if we did not build
> it in at some level?

It is partly built in, it is partly determined by
the environmental (input) history of the program.

> IMO, "its own agenda" is in the mind of the
> beholder and arises from lack of knowledge that
> would allow us to predict the robot's action.
> Once you can predict a thing's actions, the
> mystery disappears. It is then "just a machine."

If I give you a car it is your car. If I give you
an agenda it is your agenda. Of course you may
modify your car and you may modify your agenda.

When it comes to claiming that your robot has a
human like brain it is called word magic.

Give the variables labels, angry = 3, happy = 60,
bored = 20, hungry = 25, crowded = 0, lonely = 53,
surprise = 9, tired = 17 and call them emotions.

When it comes to choosing a behavior the machine
can use the emotion variables to weight the choices
along with the probability weight of a particular
choice being made anyway.

The important thing is that the machine behaves
as you might if you were sad, happy, depressed,
bored, excited and so on ...

Thus a happy robot will have a big smile on its
animation and will move about in a jolly way.
A sad robot will have an animation with droopy
eyes and will move in a slow hesitant way, or
not move at all.

It is all a bit of fun like any fiction.


John

Autonomy means doing what it does without being controlled by an other
entity.  Whether this is just a chronological list of actions to perform or
something with some intelligence built in, it still fits the basic meaning
of autonomy.

I'd argue that the more intelligence you add the "more autonomous" it
becomes.  After all, the more sophisticated a capability to deal with the
unexpected in a chronological list of actions to perform some task, the less
another entity has to step in to fix the situation.

Intelligence is, in its very most basic form, just the ability to adapt
actions based on conditions.  An iron that is tipped over and motionless for
more than a couple of seconds when hot and then turns itself off is an iron
with an "intelligent feature".  The more context, better perceptual
abilities, and planning the machine can perform, the more intelligent.  That
isn't to say intelligence is a linear phenomenon.  Stupid is, stupid does.
A dog that runs when a grizzly bear chases it is smarter than an officer and
his platoon that charge an enemy machine gunner in an open field.  Humans
are more "monkey see, monkey do" than are monkeys, actually.

Matthew



On Wed, Feb 25, 2009 at 12:17 PM, jgkjcasey <jgkjcasey@...> wrote:

>   --- In SeattleRobotics@yahoogroups.com<SeattleRobotics%40yahoogroups.com>,
> "David Wyland" <dcwyland@...>
> wrote:
> Re: Meaning of autonomous robot
>
> > What is "its own agenda?"
> >
> > 1. Where does it come from, if we did not build
> > it in at some level?
>
> It is partly built in, it is partly determined by
> the environmental (input) history of the program.
>
> > IMO, "its own agenda" is in the mind of the
> > beholder and arises from lack of knowledge that
> > would allow us to predict the robot's action.
> > Once you can predict a thing's actions, the
> > mystery disappears. It is then "just a machine."
>
> If I give you a car it is your car. If I give you
> an agenda it is your agenda. Of course you may
> modify your car and you may modify your agenda.
>
> When it comes to claiming that your robot has a
> human like brain it is called word magic.
>
> Give the variables labels, angry = 3, happy = 60,
> bored = 20, hungry = 25, crowded = 0, lonely = 53,
> surprise = 9, tired = 17 and call them emotions.
>
> When it comes to choosing a behavior the machine
> can use the emotion variables to weight the choices
> along with the probability weight of a particular
> choice being made anyway.
>
> The important thing is that the machine behaves
> as you might if you were sad, happy, depressed,
> bored, excited and so on ...
>
> Thus a happy robot will have a big smile on its
> animation and will move about in a jolly way.
> A sad robot will have an animation with droopy
> eyes and will move in a slow hesitant way, or
> not move at all.
>
> It is all a bit of fun like any fiction.
>
> John
>
>
>


[Non-text portions of this message have been removed]

Come watch Texas Instruments discuss the BeagleBoards OMAP processor, and a
few others.

http://www.justin.tv/hbrobotics


-Michael


[Non-text portions of this message have been removed]

I stumbled across this interesting technology for placing CG models
into a live video stream based on interpreting a target
graphic.  It's called Augmented Reality.  And here's a pretty
impressive example (In Japanese).

http://www.youtube.com/watch?v=oIqgrsYNKs8

Note how it works with both the camera and the target moving, and
that the model isn't constrained to sticking just to the
target.  (There's another version with the character running all over
the table, and chasing after an object held by the human).  At the
end of the video there's a demonstration of transparency, and of a
huge virtual screen.

This is all part of a program to create an artificial pop star named
Miku Hatsune.  The model also has active lip-syncing.

Of course, all technology can be turned to evil, like the Internet
meme of "Carameldansen"

http://www.youtube.com/watch?v=nQxzumHN4ps

Joshua

> OK; let's change my example a bit. [snip]
> would you agree that the
> robot is now in an emotional state?

My definition of an emotion, for the purposes of robotics, is:

An internally generated state that affects the robot's response to many
different stimuli. If it is in emotional state A then it responds to stimuli
x,y,z in one way; if it is in emotional state B then it responds to stimuli
x,y,z in a different way.

Being "sad" affects many of your responses, food tastes dull, jokes aren't
funny, you want to sleep more, etc. etc. A memory is not an emotion because
it only affects a few specific behavious.

"Angry" affects the responses to many stimuli; amonst the responses that are
affected are heartrate and breathing. "Sad" affects many responses but
heartrate and breathing are not affected.

So whether an internal state is an "emotion" depends on how you define "few"
and "many". Clearly, that's a matter of taste. Is the threshold 50%? In
humans and mammals, the difference is usually obvious. "Few" is few and
"many" is many - there don't seem to be internal states that affect "some,
i.e. more than a few but not really many" responses.


> I would argue that many utilitarian robots today are emotional.

Examples?

> Notice that we are now getting into varieties of responses: just what
> I said earlier that emotions were developed for.

Yes. But why would a robot benefit?

> Choose any of Plutchik's eight emotions (anger,
> fear, sadness, disgust, surprise, curiosity, acceptance, joy) and
> describe how you would program a robot to "be" that emotion (i.e., be
> angry, be afraid, be sad, etc.).

I find it that hard to think of a plausible program because I don't believe
it's useful for robots to have emotions. However ...

I gave the example of a warrior-bot in a platoon of warrior-bots attacking
an enemy position. The way it should behave depends on how it evaluates the
current situation. As Shakespeare says, when attacking you should "imitate
the action of a tiger". Statistically, being "brave" in that way reduces
your casualties and increases you probablility of success. You brave enemy
fire and run towards the enemy. If you believe you are going to fail and you
should preserve you hardware for future missions, you run away. In the
"brave" mode, enemies are something to run towards and shoot at, cover is
something to peer round and shoot from. In the "afraid" mode, enemies are
something to hide from, cover is something to hide behind.

That means that the stimulus of "sight of an enemy" or "sight of an rock"
triggers different behaviours. Internal "physiological" states may also be
affected - for instance, the trade off between running fast and preserving
the charge in your battery.

It seems straightforward to program such modes of behaviour. If the modes
have far reaching consequences then I'd call them emotions.

But I don't think one ought to build that sort of robot.

> if someone develops a utilitarian
> robot to protect marbles in the manner that I described, then they are
> programming an (albeit simply) emotional robot.

In your example, I didn't feel that a sufficient proportion of behaviours
were affected. But, even if they were, it's a very artificial example. Can
you think of a better one?

> we cannot program a robot to *be* anything;
> we can only program it to *do* things.

To *be* anything? What does that mean? That sounds like a philoposhical
question. We can program a robot to *be* in the "brave" state.

> if a robot cries, then it "is"
> sad (or perhaps it is crying with joy),

If the only thing it is doing is crying then it is not sad. It's peeling
onions. An emotion has to affect "many" behaviours.

I don't want to confuse "expressing emotions" with "having emotions". Humans
use facial expressions like crying to express their emotions as a part of
social communication. That is a function of our wiring. It is not a
neccessary part of being in an emotional state. Other animals have emotions
which affect many of their behaviour but they do not express them for social
reasons. An angry gorilla doesn't have an angry facial expression - an angry
chimp does. They live in different societies. A fearful zebra specifically
hides its fear from lions.

Peter

(Actually, I just invented the gorilla vs chimp example but I'd be surprised
if it wasn't true. That's what I would expect to evolve. Zebras have evolved
to look healthy and confident even when they're not. The ones that didn't
look healthy and confident were picked off.)

It's more generally useful as a way to find the relative position of a
marker/fiducial to a camera, with uses for mobile robots:  one is to
have fixed camera and a marker on top of the robot, or have a camera
on the robot and fixed markers in the environment for aiding
navigation.

Lucas

On Wed, Feb 25, 2009 at 10:53 PM, Rich Chandler <rchandler@...> wrote:
> I stumbled across this interesting technology for placing CG models
> into a live video stream based on interpreting a target
> graphic. It's called Augmented Reality. And here's a pretty
> impressive example (In Japanese).
>
> http://www.youtube.com/watch?v=oIqgrsYNKs8
>
> Note how it works with both the camera and the target moving, and
> that the model isn't constrained to sticking just to the
> target. (There's another version with the character running all over
> the table, and chasing after an object held by the human). At the
> end of the video there's a demonstration of transparency, and of a
> huge virtual screen.
>
> This is all part of a program to create an artificial pop star named
> Miku Hatsune. The model also has active lip-syncing.
>
> Of course, all technology can be turned to evil, like the Internet
> meme of "Carameldansen"
>
> http://www.youtube.com/watch?v=nQxzumHN4ps
>
>

Peter,

[PETER]
An internally generated state that affects the robot's response to
many different stimuli. If it is in  emotional state A then it
responds to stimuli x,y,z in one way; if it is in emotional state B
then it responds to stimuli x,y,z in a different way.
[/PETER]

I understood that. I'm saying that it is a probably inevitable
evolution in the development of robots which do a variety of useful
things, rather than just a single useful thing. Right now, most robots
have null responses to the majority of (potential) stimuli in their
environment. As robots become more responsive, they will become more
emotional -- by your own definition of robot emotion.

Let's use a housekeeping robot as an example. Let's say it has two
states and three stimuli, as in your definition.

state1: washDishes
state2: answerPhone
stim1: mom says "Hello, Robot," as she passed by
stim2: dog barks at door; interpretation: it wants to go outside
stim3: internal timer: put casserole in the oven

When the robot is in state washDishes, it's programming says to...
resp1: reply, "Hello, Susan. How are you, today?"
resp2: stop washing dishes; open door for the dog
resp3: stop washing dishes; put casserole in the oven

When the robot is in state answerPhone...
resp1: ignore mom (so as not to confuse the person on the other end of
the phone)
resp2: ignore dog for up to five minutes, then put person on phone on
hold; open the door and let the dog out
resp3: extend timer for up to five minutes, then put person on phone
on hold; put casserole in the oven

The robot's responses to all three stimuli are different depending on
which state it is in, so by your definition this is an emotional
robot. But would anyone characterize it as "brave" or "angry" or "sad"
or..? "Busy," perhaps, if it was on the telephone.

Best,
Joshua

Peter,

(I broke my response into two message, to keep the length of each
reasonably short. Hope you don't mind. Some ideas do cross between the
two.)

[PETER]
To *be* anything? What does that mean? That sounds like a
philoposhical question. We can program a robot to *be* in the "brave"
state.
[/PETER]

If you consider my statement in the context in which it was written,
it should obvious that I wrote it to avoid philosophical problems. I'm
not just talking about semantics; I'm talking about practical limits
of what we can instruct a robot to do. We can program a robot to
behave in a manner which is consistent with behaviors which we
associate with an emotional state, but we cannot program a robot to
"feel" or "have" or "be" (in) an emotional state without encountering
the very philosophical problems that you have said you prefer to
avoid. When you say, "if it walks like a duck and quacks like a duck,
then it's a duck," you are implicitly admitting that what a duck *is*
is in no other way inferred than by what a duck *does*.

[PETER]
I don't want to confuse "expressing emotions" with "having emotions".
Humans use facial expressions like crying to express their emotions as
a part of social communication. That is a function of our wiring. It
is not a neccessary part of being in an emotional state.
[/PETER]

I repeat: Any algorithmic description of a robot's state is
necessarily a description of the robot's behavior whilst in that
state. You cannot program a robot to "have" sadness, except to specify
what the robot does in a "sad" state, that distinguishes that state
from all the non-sad states the robot could be in. With all due
respect, when you say that you do not want to confuse "expressing
emotions" with "having emotions," it is you who introduces the
semantic problem, and so it falls to you to provide a demonstration of
a robot "having" an emotional state that is independent of the robot
"expressing" that state.

[PETER]
Other animals have emotions which affect many of their behaviour but
they do not express them for social reasons. An angry gorilla doesn't
have an angry facial expression - an angry chimp does. They live in
different societies. A fearful zebra specifically hides its fear from
lions.
[/PETER]

If an angry gorilla does not have an angry facial expression, then how
do you know it's angry? What does it *do* that signifies to you that
it is angry? I suggest that is the practical limit of what we can
engineer robotically (since you dislike philosophy, I'll leave alone
the epistemological problems of you having access to the gorilla's
internal state). If you think we can transcend that limit, please
explain how.

Best,
Joshua

Joshua,

Comments below,

> -----Original Message-----
> On Behalf Of Joshua Madara
>
> I understood that. I'm saying that it is a probably inevitable
> evolution in the development of robots which do a variety of useful
> things, rather than just a single useful thing. Right now, most robots
> have null responses to the majority of (potential) stimuli in their
> environment. As robots become more responsive, they will become more
> emotional -- by your own definition of robot emotion.

I don't see the inevitability.  Robots should have little or no response to
a lot of trivial stimuli, unless it's *their* job.  Being more responsive is
not becoming emotional; just responsive.

>
> Let's use a housekeeping robot as an example. Let's say it has two
> states and three stimuli, as in your definition.
>
> state1: washDishes
> state2: answerPhone
> stim1: mom says "Hello, Robot," as she passed by
> stim2: dog barks at door; interpretation: it wants to go outside
> stim3: internal timer: put casserole in the oven

OK, tasks are wash dishes, answer phone, mind the dog and put casserole in
oven at the appropriate time.

>
> When the robot is in state washDishes, it's programming says to...
> resp1: reply, "Hello, Susan. How are you, today?"
> resp2: stop washing dishes; open door for the dog
> resp3: stop washing dishes; put casserole in the oven

Good start; simple priorities here.  Washing dishes task can be interrupted
to greet the mistress, open the door for the dog, and load oven at the
appropriate time.  No emotion there.

> When the robot is in state answerPhone...
> resp1: ignore mom (so as not to confuse the person on the other end of
> the phone)
> resp2: ignore dog for up to five minutes, then put person on phone on
> hold; open the door and let the dog out
> resp3: extend timer for up to five minutes, then put person on phone
> on hold; put casserole in the oven

A little more complex priority structure (hierarchical), but nonetheless
priority driven.

Simple rules say no need to say hi to mom if you're on the phone (just nod
to her).  More rules to delay but still get the food in the oven.  The dog
can wait as well.  I'm not feeling the emotion!

Maybe if you had the robot get angry and yell at the dog?  ;>)

>
> The robot's responses to all three stimuli are different depending on
> which state it is in, so by your definition this is an emotional
> robot. But would anyone characterize it as "brave" or "angry" or "sad"
> or..? "Busy," perhaps, if it was on the telephone.
>
> Best,
> Joshua

I see different states, on the phone, washing, etc, and different priorities
for handling events.  Nothing more needed here.

Ever see the movie "Cherry 2000".  Quite a while ago.  If memory serves, the
plot is basically (from memory)

* Man has girl "friend" robot.  'bot relates to him, talks to him, is
interested in what he likes, seems to have emotions, etc...  (memory of
relationship)

* All the robots get collected and discarded for some reason or another.

* Man decides he *MUST* have this particular robot (memory chip of
experiences together)

* Man manages to get the memory chip, now just needs the 'bot.

* Man goes on "quest" to find a 'bot of the same model.  (good girl guide..)

* Man finally finds a suitable 'bot and is able to use the memory chip,
bringing the 'bot "back to life".

* Cherry is exactly the same as before.  Same responses to same stimuli as
before....

* Man realizes that the 'bot wasn't that great after all, once you know
exactly how it's going to respond.

* Man remembers the girl guide, a *real* girl...

* Write your own ending from there.


OK simple little "B" move, but I think the writers had some good points, and
this simple (cute) story obviously had a point.  I wouldn't be surprised if
that's the way things might go with 'bots, if we did elect to give them
"emotions".

Best regards,

Alan KM6VV

Alan,

"Simple rules say no need to say hi to mom if you're on the phone
(just nod to her). More rules to delay but still get the food in the
oven. The dog can wait as well. I'm not feeling the emotion!

"Maybe if you had the robot get angry and yell at the dog? ;>)"

1) Funny you mentioned yelling at the dog, because I almost included
the very same in an afterthought. :-)

2) Your response seems to overlook the point of my reply, which is
that I gave an example of robot behavior, that meets the definition of
robot emotion that Peter provided, but is clearly not emotional in the
sense that we commonly agree on (which is why I said, would anyone
characterize it as "brave" or "angry" or "sad"? it was rhetorical; I
would guess not). I was careful to be clear more than once that it was
Peter's definition that I was working from. Insofar as my hypothetical
robot conforms to Peter's definition of an emotional robot, saying
that my robot is unemotional is really saying that Peter's emotional
robot is unemotional, i.e., it's just pointing out the difference that
Peter and I were already having; it doesn't solve anything.

Sincerely,
Joshua

Hi John,

--- In SeattleRobotics@yahoogroups.com, "jgkjcasey" <jgkjcasey@...> wrote:
>
> --- In SeattleRobotics@yahoogroups.com, "David Wyland" <dcwyland@>
> wrote:
> Re: Meaning of autonomous robot
>
> > What is "its own agenda?"
> >
> > 1. Where does it come from, if we did not build
> > it in at some level?
>
> It is partly built in, it is partly determined by
> the environmental (input) history of the program.
>

I agree. The programmer has to put in something to start with. But as
I read the original definition, if it is built-in, it is no longer "its
own agenda."  To quote from the original email from Dr. Bruce:

Quote:
> Proceedings of the Third International Workshop on Agent Theories,
Architectures and Languages 1996 "An autonomous agent is a system
situated within and a part of an environment that senses that
environment and acts on it, over time, in pursuit of its own agenda
and so as to effect what it senses in the future."
>
> Here, however, the microcontroller based line follower fails to meet
the definition because it's doing nothing but following the
instructions in it's program - it's following the programmer's
agenda... not it's own agenda. Hence it simply isn't autonomous.
>
End Quote

To paraphrase, if a robot acts according to a program, it is following
the programmer's agenda, not "its own agenda."

However, these two paragraphs are somewhat contradictory. The first
one implies that it is possible for an autonomous agent system to have
"its own agenda," yet if it just executes a program, it does not have
"its own agenda" - it will have the programmer's agenda.

And here we are again. On the one hand, there is the wide-spread
belief that it is possible for a robot/autonomous agent to have its
own agenda; yet if it executes a program we understand, it will not
have its own agenda.

My simple interpretation is that if we program a robot and understand
how the program works, it is no longer intelligent and cannot have
"its own agenda" because it is predictable. I see this as the
unresolvable contradiction of AI.


> > IMO, "its own agenda" is in the mind of the
> > beholder and arises from lack of knowledge that
> > would allow us to predict the robot's action.
> > Once you can predict a thing's actions, the
> > mystery disappears. It is then "just a machine."
>
> If I give you a car it is your car. If I give you
> an agenda it is your agenda. Of course you may
> modify your car and you may modify your agenda.
>

The use of "your agenda" in this case is not the same as "its own
agenda" as used in the original quote.

To follow your lead, you may give me your agenda. I may accept your
agenda, making it my agenda. So far, so good. If I take a printed
copy of my agenda and put it on a table, it will still be my agenda,
not the table's agenda.

Likewise, if the agenda is a scheme for washing clothes in a washing
machine and I program the machine to execute the agenda, it is still
my agenda, not the washing machine's agenda - at least according to
common usage. This is true even though the washing machine is an
autonomous device according to the provided definition of autonomy.

If I program the robot to execute my agenda, it will still be my
agenda - in the robot. The key is that the robot is not an entity that
"receives" anything. It is a machine that executes a program that we
provide, the program that executes my agenda. There is nobody home to
"accept" the agenda.

The robot would have to be intelligent to do so, and we do not know
how to do intelligent. We do not know how to make a "somebody" in any
real sense. The best we can do is fake it. A little.

If you program an automatic machine in such a way that someone is
harmed, it is not the machine's fault, it is your fault. There is
nobody in the machine to blame. And we accept this.


> When it comes to claiming that your robot has a
> human like brain it is called word magic.
>
> Give the variables labels, angry = 3, happy = 60,
> bored = 20, hungry = 25, crowded = 0, lonely = 53,
> surprise = 9, tired = 17 and call them emotions.
>
> When it comes to choosing a behavior the machine
> can use the emotion variables to weight the choices
> along with the probability weight of a particular
> choice being made anyway.
>
> The important thing is that the machine behaves
> as you might if you were sad, happy, depressed,
> bored, excited and so on ...
>
> Thus a happy robot will have a big smile on its
> animation and will move about in a jolly way.
> A sad robot will have an animation with droopy
> eyes and will move in a slow hesitant way, or
> not move at all.
>
> It is all a bit of fun like any fiction.
>

Precisely. It is imitation emotion, like acting. Another form of the
Eliza effect (imitating human interactions).

>
> John
>

--- In SeattleRobotics@yahoogroups.com, "David Wyland" <dcwyland@...>
wrote:

> If I program the robot to execute my agenda,
> it will still be my agenda - in the robot.

And your agenda is evolution's agenda.

John

On Fri, 27 Feb 2009 08:27:37 -0000, "David Wyland" <dcwyland@...>
wrote:

> And here we are again. On the one hand, there is the wide-spread
> belief that it is possible for a robot/autonomous agent to have its
> own agenda; yet if it executes a program we understand, it will not
> have its own agenda.
>
> My simple interpretation is that if we program a robot and understand
> how the program works, it is no longer intelligent and cannot have
> "its own agenda" because it is predictable. I see this as the
> unresolvable contradiction of AI.

The contradiction doesn't have to be there.

If the steps the robot takes are directly programmed "do this, and then do
that", then I would agree, its not really autonomous.

However, if what you programmed is essentially a virtual machine, that
executes simple sequences of data, and the robot itself has assembled those
sequences (through learning), then I would say the robot can definitely
have its own agenda.

Later,
Jon

--------------------------------------------------------------
    Jon Hylands      Jon@...      http://www.huv.com/jon

   Project: Micro Raptor (Small Biped Velociraptor Robot)
            http://www.huv.com/blog

> If the steps the robot takes are directly programmed "do this, and then do
> that", then I would agree, its not really autonomous.
>
> However, if what you programmed is essentially a virtual machine, that
> executes simple sequences of data, and the robot itself has assembled those
> sequences (through learning), then I would say the robot can definitely
> have its own agenda.
I agree with Jon.

But . . . what is the point in building a mass-produced robot that has
its own agenda and doesn't do what you want it to do?

John
www.societyofrobots.com



2009/2/27 Jon Hylands <jon@...>:
> On Fri, 27 Feb 2009 08:27:37 -0000, "David Wyland" <dcwyland@...>
> wrote:
>
>> And here we are again. On the one hand, there is the wide-spread
>> belief that it is possible for a robot/autonomous agent to have its
>> own agenda; yet if it executes a program we understand, it will not
>> have its own agenda.
>>
>> My simple interpretation is that if we program a robot and understand
>> how the program works, it is no longer intelligent and cannot have
>> "its own agenda" because it is predictable. I see this as the
>> unresolvable contradiction of AI.
>
> The contradiction doesn't have to be there.
>
> If the steps the robot takes are directly programmed "do this, and then do
> that", then I would agree, its not really autonomous.
>
> However, if what you programmed is essentially a virtual machine, that
> executes simple sequences of data, and the robot itself has assembled those
> sequences (through learning), then I would say the robot can definitely
> have its own agenda.
>
> Later,
> Jon
>
> ----------------------------------------------------------
> Jon Hylands Jon@... http://www.huv.com/jon
>
> Project: Micro Raptor (Small Biped Velociraptor Robot)
> http://www.huv.com/blog
>
>

>> If the steps the robot takes are directly programmed "do this, and then
>> do
>> that", then I would agree, its not really autonomous.
>>
>> However, if what you programmed is essentially a virtual machine, that
>> executes simple sequences of data, and the robot itself has assembled
>> those
>> sequences (through learning), then I would say the robot can definitely
>> have its own agenda.
> I agree with Jon.
>
> But . . . what is the point in building a mass-produced robot that has
> its own agenda and doesn't do what you want it to do?

People continue to have children, what's the difference? :)

Ed Okerson

John wrote:
And here we are again. On the one hand, there is the wide-spread
belief that it is possible for a robot/autonomous agent to have its
own agenda; yet if it executes a program we understand, it will not
have its own agenda.

<snip>

The robot would have to be intelligent to do so, and we do not know
how to do intelligent.


Dr. Bruce replies:
I agree that if all the robot is doing is following the instructions in a
program, then it isn't autonomous - the programmer is in reality still in
control of the robot...  such a robot doesn't have it's own agenda.

However, I - along with a number of other experimental roboticists -
currently write code that gives the robot tools to do certain things....
vision, speech, memory, etc.  And we're designing programs that then use
these tools - but (and here's the difference) these programs also modify
themselves as the robot has new experiences... so, over time, the original
programs are modified by the robot....  the robot is then acting on code
that it wrote itself....  these robots are then literally developing their
own agendas based on their own unique experiences.

I guess that we're hoping that that's how you "do intelligent".    : - )

Joshua and Alan

I agree with both of you. As Joshua says, the example fits my definition of
emotions. But as Alan says, simply prioritising responses doesn't constitue
what people call an emotion.

So my definition is wrong.

If you prioritise responses then the robot will do whatever is needed to
achive its current highest priority goal. Its response to a stimulus will
depend on how that stimulus relates to the current goal. Which sounds
awfully like by definition of what happens when an emotional state has been
triggered.

But I think we'd already agreed that a goal is not an emotion. A goal is an
internal state but not all internal states are emotions.

An emotion is a "mode" of responding. Does choosing a particular goal set
the "mode". I suppose it does. If you've ever programmed such modal
behaviour you'll have found that once in a mode, you should tend to stay in
that mode. It makes the robot more purposeful and so more successful. So a
mode gains extra priority points simply by being the current mode (or goal).

But does such a mode constitute an emotion? No-one would claim it does,
would they? So my definition needs to be modified.

Any ideas?

Peter

Hi Johsua,



> -----Original Message-----
> On Behalf Of Joshua Madara
>
> Alan,
>
> "Simple rules say no need to say hi to mom if you're on the phone
> (just nod to her). More rules to delay but still get the food in the
> oven. The dog can wait as well. I'm not feeling the emotion!
>
> "Maybe if you had the robot get angry and yell at the dog? ;>)"
>
> 1) Funny you mentioned yelling at the dog, because I almost included
> the very same in an afterthought. :-)
>
> 2) Your response seems to overlook the point of my reply, which is
> that I gave an example of robot behavior, that meets the definition of
> robot emotion that Peter provided, but is clearly not emotional in the
> sense that we commonly agree on (which is why I said, would anyone
> characterize it as "brave" or "angry" or "sad"? it was rhetorical; I
> would guess not). I was careful to be clear more than once that it was
> Peter's definition that I was working from. Insofar as my hypothetical
> robot conforms to Peter's definition of an emotional robot, saying
> that my robot is unemotional is really saying that Peter's emotional
> robot is unemotional, i.e., it's just pointing out the difference that
> Peter and I were already having; it doesn't solve anything.
>
> Sincerely,
> Joshua

Yes I believe I missed the point that you were commenting on your
interpretation of Peter's spec, rather then your position.  And I had hoped
to give my slant on things.

I believe I can see ways to give at least a few emotions to a robot,
although I think I'd prefer to implement what I'd call "modifiable
behavior".

Back on the "marbles" (that was your example?), I'm thinking that rather
then have the robot get angry at unknown visitors (I think this is
approximately the example) to keep them away from the marbles, I'd want to
implement more of a "security check point" type approach.  An access list of
"known" approved visitors would be permitted access, others would not.  The
'bot keeps a cool head at all times.

It did just occur to me that showing emotions might be warranted when a 'bot
was trying to chase away an unwanted animal visitor (wolves?).  In cases
where a bot's communication with animals with no or very limited
communication skills is needed, then emotions might communicate with them.

An example where body language (emotions?) might work for a 'bot is in
communication with very young or autistic children.  This little 'bot
appears to be especially developed for just such communications.

http://hackedgadgets.com/2007/05/12/keepon-interaction-robot/
http://beatbots.org/research/

I believe the Keepon 'bot was mentioned here.  I've made a few comments on
how one might go about making one.

Best regards,

Alan KM6VV

Since we do not yet know how to make a sentient robot, our robots are still
running "our agendas".  But they can be autonomous in carrying out that
agenda.  They basically adopt our agenda.

For a robot to truly have (create) its own agenda, it would have to be
sentient; i.e., A "Johnny 5" event.  The 'bot creates its own
wiring/programming (agenda).

Below the level of sentient robots, I think there can be plenty of examples
of autonomous robot activity.  "Gnomes" if you will.

Alan KM6VV

> --- In SeattleRobotics@yahoogroups.com, "David Wyland" <dcwyland@...>
> wrote:
>
> > If I program the robot to execute my agenda,
> > it will still be my agenda - in the robot.
>
> And your agenda is evolution's agenda.
>
> John

Re: [SeattleRobotics] Re: Meaning of autonomous robot

On Fri, 27 Feb 2009 08:27:37 -0000, "David Wyland" <dcwyland@...>
wrote:

>> And here we are again. On the one hand, there is
>> the wide-spread belief that it is possible for a
>> robot/autonomous agent to have its own agenda;
>> yet if it executes a program we understand, it
>> will not have its own agenda.
>>
>> My simple interpretation is that if we program a
>> robot and understand how the program works, it is
>> no longer intelligent and cannot have "its own
>> agenda" because it is predictable. I see this as
>> the unresolvable contradiction of AI.
>
> The contradiction doesn't have to be there.
>
> If the steps the robot takes are directly programmed
> "do this, and then do that", then I would agree, its
> not really autonomous..
>
> However, if what you programmed is essentially a
> virtual machine, that executes simple sequences of
> data, and the robot itself has assembled those
> sequences (through learning), then I would say the
> robot can definitely have its own agenda.

Learning itself amounts to the programmer plus the
environment giving the robot its agenda.

We do not choose our wants and neither does a robot.

John

--- In SeattleRobotics@yahoogroups.com, "Bruce J Weimer MD"
<bjweimer@...> wrote:

> ... I - along with a number of other experimental
> roboticists - currently write code that gives the
> robot tools to do certain things ... vision, speech,
> memory, etc. And we're designing programs that then
> use these tools - but (and here's the difference)
> these programs also modify themselves as the robot
> has new experiences...

And how they modify themselves is decided by the
programmer, the result is decided by whatever
algorithm you use and the environmental (input)
history of the robot.

> so, over time, the original programs are modified
> by the robot....

Modified according to the programmer's agenda.

> the robot is then acting on code that it wrote
> itself.... these robots are then literally
> developing their own agendas based on their
> own unique experiences.

Their own unique experiences are the source of
their own agenda. It is as I wrote earlier, the
agenda is decided by the programmer when they
decide how to use the input to modify the robot's
software. The actual agenda is then a result of
the programmers learning algorithm and the input
history of the robot.

> I guess that we're hoping that that's how you
> "do intelligent". : - )

That depends on the programmer's algorithms.

Environmental modification of a robots agenda
as determined by the programmer is not itself
a guarantee of an intelligent result.

John

Hi Peter,

I don't have a better definition yet, but I'm inclined to think of emotions
as "emotional baggage" to a particular internal state.

Militaries have "Response Conditions" or states that indicate how high the
current threat level is (loosely stated).  Perhaps something along these
lines could be applied to robotics.  Not that I want to see all of our
robots engaged in a war machine (I don't want to invoke flames either).

For a 'bot it might be more of a "civilian response condition" (but for
'bots).  So maybe there's a state for normal day to day, and one for the
"House is on fire/flood/etc", and one for "company coming" (need extra level
of clean/neatness?).  You get the picture.  Just more modes or mode
modifiers.  These are alterations of priorities.  The dog stays out if
company is coming.  Extra process needs to be run; get out the pictures of
relatives...

Back to the emotions, various emotions could be selected for imitation based
on an emotion setting (state), and would vary depending on the mode the 'bot
was in. the appropriate emotion would be presented by the 'bot.  This would
be done possibly in much of the same way as the random text of an
entertainment 'bot might make comments (if someone wanted all that).

I did come up with a possible reason for a 'bot to exhibit or imitate an
emotion.  Back to the "interfacing" with animals I mentioned in my previous
post; I could see a "backyard security" 'bot exhibiting anger (being in an
offensive vs. defensive mode) towards an unwanted animal visitor (stray cat,
skunk or raccoon; be careful not to offend a skunk!).  OK, not harm the
animal, but a little bit of offence maneuvers to dissuade a skunk or
unwanted cat from sticking around could be useful!  I don't think it would
serve any purpose to have a scary face on a CRT or LCD for the unwanted
visitor to look at.  Loud noises could help (but who wants a noisy 'bot).  A
legged "spider" 'bot could always wave its front two legs at the intruder a
la a Lynxmotion hexapod (videos on youtube!

Alan KM6VV

> -----Original Message-----
> On Behalf Of Peter Balch
>
> Joshua and Alan
>
> I agree with both of you. As Joshua says, the example fits my definition
> of
> emotions. But as Alan says, simply prioritising responses doesn't
> constitue what people call an emotion.
>
> So my definition is wrong.
>
> If you prioritise responses then the robot will do whatever is needed to
> achive its current highest priority goal. Its response to a stimulus will
> depend on how that stimulus relates to the current goal. Which sounds
> awfully like by definition of what happens when an emotional state has
> been triggered.
>
> But I think we'd already agreed that a goal is not an emotion. A goal is
> an internal state but not all internal states are emotions.
>
> An emotion is a "mode" of responding. Does choosing a particular goal set
> the "mode". I suppose it does. If you've ever programmed such modal
> behaviour you'll have found that once in a mode, you should tend to stay
> in
> that mode. It makes the robot more purposeful and so more successful. So a
> mode gains extra priority points simply by being the current mode (or
> goal).
>
> But does such a mode constitute an emotion? No-one would claim it does,
> would they? So my definition needs to be modified.
>
> Any ideas?
>
> Peter

John wrote:
Environmental modification of a robots agenda
as determined by the programmer is not itself
a guarantee of an intelligent result.


Dr. Bruce replies:
I absolutely agree!  Nothing about the methods that I've outlined guarantees an
intelligent result.

Although, parenthetically, I can guarantee that so far it's producing an
interesting result.  ; - )

The real point though is that it doesn't preclude an intelligent result.  And by
building robots like these and by programming these higher-level architectures
into them so that they can learn from their experiences, we're hoping to learn
how to actually make intelligent robots.


[Non-text portions of this message have been removed]

Alan

Thanks for the examples. I think they're the best I've heard.

My original question was "Why?". I think it's easy to add emotional states
to a robot (compared with difficult stuff like dynamic walking). And I think
it's easy to express emotions (which may or may not correspond with an
internal emotional state).

I couldn't see why it would be useful for a robot to communicate an
emotional state ("real" or "fake"). But I hadn't considered communicating
with animals. Of course, it matters whether the animal is a social animal
that cares about the emotional state of its associates (like a dog) or
doesn't (like our cat). I don't know about skunks' social lives but I guess
a skunk will run away from a big, fast, loud, scary machine that's waving
its arms and advancing. But it will do so without considering the emotional
state of the machine. A dog however, will understand the emotional state of
a approaching dog based on relatively subtle clues - the stiffness of the
gait, the curl of the lip, the position of the tail. (Which is, I think,
contrary to what you said. Expressing emotions is useful only when
communicating with an animal with very _good_ communication skills.)

So, expressed emotions may be a useful way of communicating with animals.
But I still don't really get it for humans.


As for actually _having_ an internal emotional state, I couldn't see the
point of that. All along, I'd been assuming that the state was internally
generated. I couldn't see why one would want a robot with that degree of
autonomy. But, as you say, there's no reason why the robot shouldn't have a
knob on its forehead. (I believe such a character appears in Judge Dredd).

A security-bot must always be evaluating the theat level of every input. Its
current "Response Condition" sets the threhold of what is deemed to be a
threat. That's as close to an emotion as you'd want in a robot. I guess I'd
call it a mild emotion. That's definitely not the same as a prioritised
goal.

I think most people would feel that emotions imply a less than rational
response One would always want a robot to behave rationally. In a highly
charge emotional state, people often do things that are not in their best
interests. You wouldn't want your security-bot to go berserk and injure
bystanders.

(Actually, the "losing it", "red mist" state of complete violent rage is not
irrational. It turns out to be a good strategy. Imagine you are in jail and
people as continually probing your weaknesses. They want to know how far
they can push you before you react. How far they can push you; how much they
can get from you. "Rationally", you ought to have a fixed threshold. But
that means that people will always push you right up to that threshold. You
will constantly be forced to react to threats. You better strategy is to
occasionally over-react to a minor insult, to completely lose your
self-control and beat someone to pulp. That way, no-one knows where your
threhold lies and will be careful not to approach it too closely. It's good
to have the reputation of emotional instability and irrationality.)

Peter

Hi Peter,

I had fun thinking about the examples.

> -----Original Message-----
> On Behalf Of Peter Balch
>
> Alan
>
> Thanks for the examples. I think they're the best I've heard.
>
> My original question was "Why?". I think it's easy to add emotional states
> to a robot (compared with difficult stuff like dynamic walking). And I
> think
> it's easy to express emotions (which may or may not correspond with an
> internal emotional state).

Yes, I still think adding emotional states to a 'bot is questionable.

> I couldn't see why it would be useful for a robot to communicate an
> emotional state ("real" or "fake"). But I hadn't considered communicating
> with animals. Of course, it matters whether the animal is a social animal
> that cares about the emotional state of its associates (like a dog) or
> doesn't (like our cat). I don't know about skunks' social lives but I
> guess

Of course you don't want to antagonize a skunk too much!  In the few times
I've had the opportunity to observe them; they appear to just wander about,
with little care for anything going on.  Like they KNOW other animals won't
attempt to bother them (observations without any threats around).  Dogs
would seem to be the most controllable.  Cats, as you say, don't care what
you think.  They may also learn after a few encounters that the robot won't
actually HURT them, and then will simply disregard them.  Sounds like an
interesting experiment.

> a skunk will run away from a big, fast, loud, scary machine that's waving
> its arms and advancing. But it will do so without considering the
> emotional
> state of the machine. A dog however, will understand the emotional state
> of
> a approaching dog based on relatively subtle clues - the stiffness of the
> gait, the curl of the lip, the position of the tail. (Which is, I think,
> contrary to what you said. Expressing emotions is useful only when
> communicating with an animal with very _good_ communication skills.)

I don't remember exactly what I said about dogs, but I certainly agree with
your assessment.

> So, expressed emotions may be a useful way of communicating with animals.
> But I still don't really get it for humans.

Mostly because we have much better ways of communicating.  Makes me wonder
about (no flames please) "Rap" performances.  The verse is terse and
typically doesn't convey much richness in speech.  There appears to be no
melody to the music.  It would seem that extensive jesters are needed to get
the points across.  (No offence meant to any one or group).


> As for actually _having_ an internal emotional state, I couldn't see the
> point of that. All along, I'd been assuming that the state was internally
> generated. I couldn't see why one would want a robot with that degree of
> autonomy. But, as you say, there's no reason why the robot shouldn't have
> a
> knob on its forehead. (I believe such a character appears in Judge Dredd).

I certainly don't want a robot with emotions as my "lab assistant" or
housekeeper!  And although I've explored the idea of a "watch robot" for the
backyard, it is still to be proven whether it would be any more effective
than a 'bot that simply chases off animal intruders, or sounds the alarm for
human ones.

What did you think of the use of the Keepon robot (animatron?) in
communication with autistic children?

> A security-bot must always be evaluating the theat level of every input.
> Its
> current "Response Condition" sets the threhold of what is deemed to be a
> threat. That's as close to an emotion as you'd want in a robot. I guess
> I'd
> call it a mild emotion. That's definitely not the same as a prioritised
> goal.

Yes, I'd say the elevated response condition merely modifies the existing
goals.

> I think most people would feel that emotions imply a less than rational
> response One would always want a robot to behave rationally. In a highly
> charge emotional state, people often do things that are not in their best
> interests. You wouldn't want your security-bot to go berserk and injure
> bystanders.

Certainly not!

> (Actually, the "losing it", "red mist" state of complete violent rage is
> not
> irrational. It turns out to be a good strategy. Imagine you are in jail
> and
> people as continually probing your weaknesses. They want to know how far
> they can push you before you react. How far they can push you; how much
> they
> can get from you. "Rationally", you ought to have a fixed threshold. But
> that means that people will always push you right up to that threshold.
> You
> will constantly be forced to react to threats. You better strategy is to
> occasionally over-react to a minor insult, to completely lose your
> self-control and beat someone to pulp. That way, no-one knows where your
> threhold lies and will be careful not to approach it too closely. It's
> good
> to have the reputation of emotional instability and irrationality.)
>
> Peter

Jails must be a "society turned upside down".  Rational behavior is
apparently ineffective, and a more primitive culture prevails.  Hopefully we
won't ever have to experience such.

Alan KM6VV

I have been looking for vehicle drive formulas so I can spec the
motors on my wheelbot project.
I found some sources but I think some important parameters are missing.
It seems to me that a 300 pound estimated machine that runs 10 mph on
13" tires would need a tad more torque than 30in/lbs to achieve this
feat, If anyone here would like to share their formulas I would
appreciate it.