Re: What did that thread indicate?

"Glen M. Sizemore" <gmsizemore2@xxxxxxxxx> wrote:
> <humiguel@xxxxxxx> wrote in message
> news:1127606995.524536.238130@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

> >>One thing is certain, using "shaping" (differential reinforcement of
> >>successive approximations) animals will acquire behavior that would
> >>never have otherwise been observed.
> >
> > How can you ever be sure of that?
>
> Well, you can't be completely sure, of course. But there are responses
> that you can shape that are highly unlikely. For example, you can train
> rats to press a lever that requires more static force than their body
> weight.

Humans and animals are trained to do what they do. If you train a human to
speak English, they will speak English. If you put them in an environment
were they are punished whenver they make any sound that sounds something
like some part of English, they will NEVER speak English.

If you put them in an environment that has no English and they are never
rewarded for speaking an English-like langauge, the odds of them every
devloping a full set of English-speaking behaviors on their own is so low
as to be safe to call it zero (just like the odds of someone raised in a
non English speaking environment is never expected to develop English on
their own if they are never exposed to it).

It's becaue the odds of a small behavior may not be bad (like the odds of
making a sound that sounds like the word "I" is pretty good for example).
But the more complex the behavior, the more of these unlikely events which
are required to be "shaped" into the normal behavior of the person. To
speak (and understand) English, we must have a very large set of beahviors
shaped into us. The odds that all these behavior would happen without
being lead by an environment actively trying to shape the behavior into us,
will quickly reach astronomically low values.

So the odds of a behavior happening spontainously is a function of how
close that behavior is to the current normal behavior of the animal or
human. So if speaking the word "Hey" and speaking the word "Bob" is
already part of the normal behavior of a person, then the odds of creating
the combined behavior of speaking "Hey Bob" is obviously much higher, than
the odds of someone who has never been trained to produce either of the
simpler behaviors.

The point of long term "shaping" is to transform us from one set of natural
behaviors, to some other set of natural behaviors, one small step at a
time. The space of all behaviors we can be taught to perform is
astronomically large. The odds of us reaching any new configuraion of
learned behaviors on our own, without major influence from the
envirornment, is basically zero. We reach the new set of behaviors because
the environment pushed us that way.

--
Curt Welch http://CurtWelch.Com/
curt@xxxxxxxx http://NewsReader.Com/
.

Relevant Pages

  • Re: What did that thread indicate?
    ... > speak English, they will speak English. ... > If you put them in an environment that has no English and they are never ... Well, shaping is certainly involved. ...
    (comp.ai.philosophy)
  • RE: Mulitple Language User Interface
    ... For the English XP OS, ... you can change the environment to the user`s native ... I haven`t found the MUI version for non-English OS yet, ... >> I know for the English Windows system, you can download ...
    (microsoft.public.windowsxp.general)
  • Re: scim chinese input under non GB2312 locale
    ... What I tried to find out is how to start editing Chinese on English ... IIRC on FC2/FC3 to input Chinese on English locale I have to run first; ... When I login my environment is set up as shown here: ...
    (Fedora)
  • multiple language interfaces
    ... so if one person logs on the window ... environment is spanish. ... I heard XP English addition can ... English instead of Japanese? ...
    (microsoft.public.windowsxp.general)
  • Re: DSL question.
    ... how does DSL interfere with the phone line? ... daytime and early evening who actually speak English as a first ... Americans toward people for whom English is a second language -- well, ...
    (rec.arts.tv)