Temporal_Intelligence = digital/binary system ?
cg...@onwe.co.za
As one who's read/studied AI for some decades I found my self at:
http://home1.gte.net/res02khr/AI/Temporal_Intelligence.htm
The claims resonate as mostly true.
> One of the most powerful aspects of the temporal approach to
> intelligence is that it is based on a single unifying concept: the
> relative arrival times of discrete signals.
And on deeper thought even obvious:
* in signal engineering we talk about transforming the frequency
domain to the equivalent time domain signal.
* although you claim to play-down the info-representation aspect
of AI; the digitisation (to binary form) of any/all info consists merely
of "the relative arrival times of discrete 'spikes' ". In this most simplified
case, a spike or not a spike (ie. binary).
> The only thing that distinguishes one spike from another is its time
> of arrival.
Can be transformed (with equivalence) to:
"The only thing that distinguishes one discrete 'time slot' is its
contents or not of a spike/bit ".
-- Chris Glur.
tedd
>
> And on deeper thought even obvious:
> * in signal engineering we talk about transforming the frequency
> domain to the equivalent time domain signal.
Well... you don't.
When you go from one domain to the other, you are making the assumption
of what you get is what there was -- it isn't.
For example, take a 10Hz signal at zero degree phase shift and another
10 Hz signal at a 180 degree phase shift -- what' the end result?
Nothing but a straight dead line. Now, do a FFT on that signal and tell
me what created the signal -- you can't.
The same goes if I phase shift the two signals by any amount -- you
can't tell me what the components were to create the resultant signal.
You can only tell me what frequencies/phases needed to represent the
known signal.
So, the point I'm trying to make is that transforms are not
representative of what's really going on. They are only tools that we
can use to help us make decissions. But, we should know their limitations.
tedd
Rick Craik
<cg...@onwe.co.za> wrote in message news:ag0j7n$b6v$1...@ctb-nnrp2.saix.net...
> Hi,
> As one who's read/studied AI for some decades I found my self at:
> http://home1.gte.net/res02khr/AI/Temporal_Intelligence.htm
>
> The claims resonate as mostly true.
>
> > One of the most powerful aspects of the temporal approach to
> > intelligence is that it is based on a single unifying concept: the
> > relative arrival times of discrete signals.
The term "bit" is a short form for "binary digit", is this what you
meant by a "digital/binary system"? I find the term "bit" most
useful when describing a communication channel capacity
of a system. I have no problem with "Temporal Intelligence"
describing it's communication system in a digital binary notation
(spike in an axon).
I find it is misleading to describe an entire intelligence system
in such a notation that classifies it as a "digital/binary system".
For example; "relative arrival times" suggests that a binary 0 might
not have the same meaning as implied in Boolean logic. I may resort
to a unary digital notation, and a variant of Boolean logic to describe
the intelligence system.
Regards,
Rick
Traveler
>Hi,
> As one who's read/studied AI for some decades I found my self at:
>http://home1.gte.net/res02khr/AI/Temporal_Intelligence.htm
>
>The claims resonate as mostly true.
>
>> One of the most powerful aspects of the temporal approach to
>> intelligence is that it is based on a single unifying concept: the
>> relative arrival times of discrete signals.
>
>And on deeper thought even obvious:
>* in signal engineering we talk about transforming the frequency
> domain to the equivalent time domain signal.
>* although you claim to play-down the info-representation aspect
> of AI;
It is not a claim. From the point of view of an observer looking in,
spikes do represent specific events. But from the point of view of the
brain (that is the only point of view that matters if you're
interested in understanding how the brain learns), they are just
discrete signals. Look at it this way, if the brain already knew what
the spikes symbolized, what would be the purpose of learning? Even
symbols are learned. What symbolizes a symbol?
> the digitisation (to binary form) of any/all info consists merely
> of "the relative arrival times of discrete 'spikes' ". In this most simplified
> case, a spike or not a spike (ie. binary).
It's more than that. A spike may arrive 10 milliseconds after another
or 1 second. The relative time of arrival is what is important. Your
binary analogy seems somewhat atemporal to me.
>> The only thing that distinguishes one spike from another is its time
>> of arrival.
>
>Can be transformed (with equivalence) to:
>
>"The only thing that distinguishes one discrete 'time slot' is its
>contents or not of a spike/bit ".
I think you may be missing an important point here. What gives the
time slot its relevance is not whether or not it is occupied but
whether or not the system can identify all the occupied slots
originating from the same sensory pathway and find temporal
correlations between those slots and those belonging to another
sensory pathway. Again, it all comes down to the relative times of
arrival.
Regards,
PS. As an old AI hound, you are making a mistake developing an
interest in my site. The GOFAI community (at least the ones who know
about me) don't want to have anything to do with me or my ideas.
You've been warned. :-)
The Silver Bullet:
http://home1.gte.net/res02khr/AI/Reliability.htm
Temporal Intelligence:
http://home1.gte.net/res02khr/AI/Temporal_Intelligence.htm