Subject:
|
Re: Robots In Love?
|
Newsgroups:
|
lugnet.robotics
|
Date:
|
Tue, 1 Feb 2000 03:57:12 GMT
|
Viewed:
|
1478 times
|
| |
|
|
Hi Michael,
Wow! I never really expected anyone to read it! Thanks!
I've read Minsky's "Society of Mind" and was generally unimpressed with it.
It's not so much that I think anything he said was wrong as much as it seemed
dated to me (the book was written in 1985). While you'll definitely see
similarities between what I do and Minsky's societies of agents, I believe
there's too much functional decomposition in his approach, and that's
definitly NOT the direction I want to head.
As for the notation of Love, yes, it's a bit difficult. But physics uses a
similar notation, where there's a subscript, as in Xy where y is a subscript
and represents something along the lines of velocity or mass. In fact, Love is
really just a formalization of that approach. In Love you'd write Y^x. I may
one day change the notation to match what physics uses. In the work I plan on
doing with robots and Love, I'll be stripping away the superscript entirely,
because I'm only interested in what the base numbers represent.
Continuum jumps don't play a major role in anything I've done so far and
perhaps should be moved to the "Tools" section, with functions being moved out
of the "Tools" section and getting it's own section. The jumps specify
identities, equalities to be exact, but could be used for inferances as well.
I'm glad to hear your feedback on them. I never really thought about the
combinatoral explosion problem they could represent because, as I said, I
don't use them that often. I don't see a way around this though, as it seems
to me to be a product of our minds producing so many different ways to express
the same thing, rather than a product of notation. Do you have any ideas on
the matter?
The goals I have with using Love is really much smaller than something along
the lines of Minsky's agents. They would be parts of agents in his method. I
see the simulated emotion that I can produce with Love being useful as "tags".
I'm taking the tags concept from John Holland's book on Complexity, which is
called "Hidden Order". Tags work in Complexity the way icons work in a GUI OS,
like Windows or the Mac. They are simple things that represent much more
complicated ideas.
As an example, look at a simple cruise and avoid program based on the
Subsumption Architecture. In these type of programs there are "chunks" that
say "GO", "TURN LEFT" and "TURN RIGHT". They work, but all three could be
suplimented with an emotional tag. Why do you want to GO? Why do you want to
TURN LEFT? Adding a tag like "LUST" or "FEAR" adds an enourmous amount of
informational content to the request. I want to GO because the data I've been
programed to LUST for is in front of me. I want to TURN LEFT because the
darkness I've been programmed to FEAR is to the right. With this type of
information, the decision making section of the Subsumption Architecture can,
I hope, make better decisions.
Another example would be using BOREDOM. A robot that examines its light
sensors for an extended period of time and detects no changes could become
BORED. Enough BOREDOM, and the Subsumption Architecture decides it's time to
do something about it. This may play a role in error handling for map
following routines, or when robots simple get stuck in a corner.
What Love gives me is the ability to represent these tags using only a single
bit of RCX memory. If you take a look at the column charts, you'll see the
columns represent the numbers 1, 2, 4, 8, 16, 32, etc. They translate
perfectly into bit flags. It's as compact as you can get, and on the RCX,
that's what I need.
http://alive0.tripod.com/LoveIntro.htm
David Leeper (really is pleasantly suprised you read the site!)
In lugnet.robotics, Michael Naunton writes:
> Ok, I read it. The notation is extremely strained, and redefining X^Y
> to mean Y*X' (where X' is the underlying concept representing X) does
> not help. On the plus side, at least you figured out that orthogonal
> concepts need to be represented as a series of powers of two if the
> whole system is not to immediately break down. The "continuum jumps"
> are a weak way of specifying inferences or identities (i couldn't tell
> which from the examples,) and suffer from combinatorial explosion.
>
> I'd suggest picking up a copy of Minsky's "Society of Mind" or Schank's
> "Scripts, Plans, Goals, and Understanding." These both deal with
> high-level symbolic information processing.
>
> Regards,
> Michael Naunton
>
>
> David Leeper wrote:
> >
> > Hi All,
> >
> > I've just put up a new web site. It's about a new type of mathematics I
> > invented about a decade ago. It's called Love, and works well at
> > mathematically representing emotion. What does this have to do with robots?
> > Well, I AM building an android.
> >
> > Here's the URL:
> > http://alive0.tripod.com/Love.htm
> >
> > David Leeper (wouldn't deprive Adam of a heart)
|
|
Message has 1 Reply:
Message is in Reply To:
 | | Re: Robots In Love?
|
| Ok, I read it. The notation is extremely strained, and redefining X^Y to mean Y*X' (where X' is the underlying concept representing X) does not help. On the plus side, at least you figured out that orthogonal concepts need to be represented as a (...) (24 years ago, 1-Feb-00, to lugnet.robotics)
|
4 Messages in This Thread:
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
