I'm becoming increasingly horrified by how people are entertaining
Mentifex. We really really don't have any more time to waste on that
crackpot asshole. =|

We are right on the eve of the singularity. We have no more time for
bullshit.

In a desperate effort to steer things back on course, here is a list of
things, from my POV, from my limited understanding, that we need to do
to get to AGI.


1. Our current neural models are fairly good but there is a major trick
that they seem to be missing: The space-time rotation. For example, in
the hearing system, the ear translates the temporal frequency
information into a "tono-topic" map on the auditory cortex. This is a
time -> space rotation of the signal. I'm pretty sure the visual system
rotates the spatial information from the eyes into a temporal signal
that is used to detect patterns. In computer vision, the latest
technique I know of, hack-botches this by scanning the perceptron across
the input. This is a hack! Rotate it into a temporal signal and it
becomes much easier to analyze.


2. There is quite a bit of research into evolving better deep networks,
tweaking the number and characteristics of deep networks to try to
achieve various metrics. THIS IS WRONG. WRONG!!!! I SAY...
Wroooooong!!!!!!!!!!!!!!   What the brain does is have a very small
number of stereotypical neural circuits that it trains on different
inputs/behaviors and recruits them as needed. Search for the
cortico-thalamo-cortical loop.


3. >>>> Perception is imagination <<<<. -> figure out how to make the
above powerful enough to produce an imagination process sufficient to
make high quality short-term (on the order of a fraction of a second)
predictions of the input signal.


4. A robotic and/or virtual avatar system so that the AI can experience
a reasonable approximation of humanness to facilitate psychological
development, communication, and education, it may not be strictly
necessary but it will make it much much easier for sub-geniuses to
develop and use AI systems.


Ideally, there would be more powerful self-optimization processes to
tune the dimensionality of the neural matrices and such, but that is not
really necessary at this point to reach human equivalence.





-- 
Please report bounces from this address to [email protected]

Powers are not rights.


------------------------------------------
Artificial General Intelligence List: AGI
Permalink: 
https://agi.topicbox.com/groups/agi/T507c404b4595c71c-Mc18f99aa4efccc0e4c49eedc
Delivery options: https://agi.topicbox.com/groups

Reply via email to