[ExI] Neural networks score higher than humans in reading and comprehension test

spike spike66 at att.net
Wed Jan 17 01:05:23 UTC 2018



From: extropy-chat [mailto:extropy-chat-bounces at lists.extropy.org] On Behalf Of John Clark


>…I don't know why people make such a big deal about Shakespeare, all he did was place one ASCII character after another, even a child can do that.    John K Clark  



Ja!  All Shakespeare did was take a bunch of famous quotes and string them together to make a story out of them.  Child’s play.



>… I look at something marvelous but so complex I can't understand it, ​I then ​break that marvelous thing up into smaller and smaller parts ​until​ eventually I come to a part that is so simple it can easily be understood


Let us this about chemistry for a minute.  Big complicated field of study, churns out lots of PhDs and such.


But let’s take a single hydrogen atom and think of how many parameters we need to describe that atom completely.  We need a velocity with respect to something, the excitation state of its electron (if present) any neutrons, etc, but we can see that it is a finite number of parameters needed to completely describe that atom.


Likewise we can do the same trick with a helium, with a lot more numbers required, but if we have an ordered understanding of what each of those numbers mean (electronegativity, mass, phase change conditions and forth) we can completely describe that helium, and the same argument can be used to describe every element in the chart, well over a hundred of them.


A list of numbers is a vector from the point of view of the computer; it doesn’t care how many numbers are in the vector.  It handles multi-dimensional space with no sweat on its processor.  It’s what computers do.  No worries.


The vector completely describing everything there is to know about an atom can be used to completely understand how two atoms can interact with each other, with some kind of operator.  So with that, any two-atom molecule can be completely described, and so can molecules with three atoms, or four or more, with each compound made of molecules which can be completely described by a list of numbers, or a vector (for our linear algebra fans among us.)  


By logical extension, all chemistry could be reduced to a field of mathematics, but I prefer to think of it as elevated to a field of mathematics.  The computed doesn’t need to memorize a bunch of chemical names.  It doesn’t care if this list of numbers is a molecule of paradichlorobenzene, it knows everything that molecule can do, knows the location of every reaction site, knows how it interacts with other similar molecules and so on.


Structures are made of molecules that interact with each other in a special way, so these too can be described as vectors (really really big ones, but still just a list of numbers, the kind of things computers do so very well.)


Living tissues are made of structures, and membranes are made of tissues and cells are made of membranes, so if we can completely describe interactions of molecules with vectors, we can operate ourselves right on up to the cell level with vectors and from there on up to organs, such as… our brains.


Sure it is soft and squishy, but it is made up entirely of atoms which we already agreed could be described by a list of numbers.  It is a machine that runs on the principles of chemistry.  Given sufficient technology, chemistry can be modeled (we are doing that today already.)  With still more technology (more memory and processor power) then more complex chemistry can be modeled or simulated.


If we ever get to the point where we can sim a brain… we have arrived.


We don’t know if we are smart enough to ever sim a brain.  Can a brain simulate a copy of itself?  We don’t know that.  But unless we assume a brain is smart enough to create a simulation of itself, we never will.  I hope we will.


So… I prefer to think we eventually will, and keep working toward that goal.  Because if we can’t, we already know what will become of us.  If we make the goal, we don’t.  I want to not know what will eventually become of us.  So… I am thrilled with every cool new achievement, even if we know it isn’t “understanding” or it isn’t “intelligence” exactly but just a cool math trick.


Fun parting shot on my little essay: I have been following AI for many years.  I was one of those caught up in the hype in the 1980s over neural nets, and tried my hand at it back in the day.  I realize we ended up with a big goose egg on our time invested, but just in the last year or two, notice how it has been one thing after another?  Even if it is misleading headline grabbers (and I do think most of it is that) it might be something new and something promising.  


Is this a great time to be living or what?











-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20180116/8aa23292/attachment.html>

More information about the extropy-chat mailing list