Re: Temporal Learning

On 11 Oct 2005 18:50:31 GMT, curt@xxxxxxxx (Curt Welch) wrote:

>Traveler <traveler@xxxxxxxxxx> wrote:
>> On 10 Oct 2005 21:27:22 -0700, "feedbackdroids"
>> <feedbackdroids@xxxxxxxxx> wrote:
>
>> >Obviously, from an evolutionary perspective, it's better for humans to
>> >have mechanisms which allow for great flexibility and adaptation.
>>
>> What is obvious is that it would have been much better for survival to
>> be born with pre-wired knowledge about one's environment.
>
>Well, what's obvious is that it would have been better to have both. Lots
>of pre-writed knowldege so we don't have to waste 20 years in school, but
>also lots of flexibility and adaptation. We probably have a lot of both as
>it is.

I think that humans inherit of a lot less knowledge than they learn
over a lifetime.

>I don't tend to argue with Dan that we have lots of nature in is, even
>nature which is missing from apes. I just argue that the key to AI is not
>figuring out all that nature but instead, figuring out the underlying
>principle about how the adapation works so well in both apes and humans.

I agree.

>An IBM PC of today, has lots of modules and stuff in it that you won't find
>in a computer from 1955. Today's PC is far more advanced than what we had
>50 years ago. But, spending endless hours studying the current Xeon
>chipsets does little to help solve the mystery of the computer when we
>still don't understand the basics of what a stored program computer is and
>how it works. Until we understand how that computer from 50 years ago
>works, looking at all the advanced stuff in a current PC only makes the job
>of identifying the underlying concepts harder because there's so much more
>stuff we have to ignore.

The basic principle (executing a sequence of instructions) used in
making and programming computers has not changed (evolved) since the
days of Babbage and Ada Lovelace. Only the capacity and speed have
changed.

>John likes the idea of studying the tree of evolution to see how we got to
>where we are. In that path, we will no doubt uncover all the mysteries. I
>think however that the important step happaned around the time of the
>neocortex showed up. And if you spend time studying life before the
>neocortex, you will understand more about the brain, but not much about our
>strong powers of adaptation which I think is the key missing link here.

There is no reason for the neocortex to evolve, IMO. The vast majority
of successful lifeforms (insects, bacteria, etc...) on earth don't
have one and they all survived for billions of years. There is no
pressure for them to change. In fact, after billions of years, they
still don't have a neocortex. I am not saying that the neocortex did
not arrive late on the scene. I am saying that there was never any
evolutionary pressure for it to arrive. Some other causal influence
was responsible. It is evolutionary but not in the Darwinian sense.

>Some people don't seem to believe thare are any important underlying
>principles which allow all this to work (or that we understand all of them
>already). They feel the only thing left to do is to figure out module by
>module what all the parts do in a human brain.
>
>I just happen to believe there are still some important underlying
>principles we have not yet mastered, and until we master those, we never
>will really understand what we are looking at when we dig into the brain.

Certainly.

>I'm not really interested in duplicating humans anyhow. I just want to
>understand the technolgy that makes us work, so I can adapt it and use it
>in all the very non-human machines I want to build - like a vending machine
>that you can talk to as if it were human, or a help desk machine that could
>answer questions, do research, and serve as a communication hub between
>humans. Or a robot smart enough that I could teach it to take out the
>trash, or replace burnt out light bulbs in the house, or cut my grass for
>me, and pull weeds.

Yes. IMO, if we only understood the principles of learning and
adaptation, we would achieve amazing wonders with the computing power
of our current desktop systems.

>> This is not
>> observed in humans. You can believe in the magic of evolution all you
>> want but, IMO, evolution requires a lot of samples and opportunities
>> for combining genes. The problem with that is that opportunities are
>> extremely rare in a small population which was the case in the
>> beginning.
>
>Population size doesn't really prevent it from working, it just limits the
>speed of adapation to a changing envrionment.

It prevents it from working the way we've been told. In order for new
species to emerge, there must be a lot of opportunities for mutations.
The problem with the mutation hypothesis is that, the more complex the
organism, the less chance any mutation has to be successful. The
mutation process discourages complexity. Big time.

>> We should be seeing a lot more new human species emerging
>> now (we have billions of mating specimens) than in the distant past
>> when populations were sparse.
>
>No we shouldn't be. Evolution is not a processes of creating new species.

You're mistaken. Darwinian evolution is precisely about the origin of
species.

>It's a processes of finding the optimal solution to the problems created by
>the environment. It's the environment that defines the species, not the
>processes of evolution. When the environment changes, the species change
>to adapt.

But there is no change, genetically speaking. Unless the gene that
allows an individual to survive an environmental hardship already
existed in the individual, the individual has no chance of surviving.
Nothing new is created. Obviously the genes were already present in
some individuals. Why? Where do they come from? How did they emerge
without previous evolutionary pressure? Pressure from the environment
does not create new genes. The genes are already there. Why? That's
the chicken and egg problem of Darwinian evolution. It needs
environmental pressure to change but it cannot change unless the genes
somehow existed before the pressure arrived to cause the change.
Mutation alone cannot account for new gene creation because the more
complex the organism the less successful the mutations. In the
Darwinian scheme of things, complexity is its own worst enemy.

Let me add that natural selection is kind of like reinforcement
learning. It does not create new genes; it only selects pre-existing
genes via the mechanism of sexual breeding. Likewise, RL does not
created new behaviors; it selects existing behaviors; some other
principles create the behaviors. It's a selection process.

>The only way to force variation in a species, is to split then into
>separate breeding pools and put each pool into a separate environment.
>That allows the breeding pool to evolve into a different form optimized for
>the needs of their unique environment. This is what happaned to man, as he
>spread around the world and became isolated in different areas with
>different environmential demands.

If other species are any indication, we should witness several species
of human-like creatures that are more or less equally intelligent even
though they cannot interbreed. This is not observed and, regardless of
claims to the contrary, the fossil record does not show evidence of
intelligent humanoids who did not interbreed. There are all kinds of
bird species who cannot genetically interbreed. Same with lizards,
spiders, fish, etc...

>> This is not observed. On the contrary,
>> the trend is toward a single homogeneous race.
>
>Of course it is. We are cross breeding at increasing higher rates. That
>means we are forming one large breeding pool for the entire environment of
>earth and we are becoming a species which is optimized for one large single
>environment - Earth. Do you want to see man evolve apart, send a million
>people to Mars and don't let the Martians cross breed with the Earthlings
>for the next 100,000 years and then see what happens.

A billion years from now, they would still be human.

>Man is not an individual. He's part of a large breeding pool which works
>as a single large machine designed for the purpose of survial by adapting
>to the environment. If you want to see these large machines take different
>paths, you have to cut them in half and create two machines and then left
>them each go off on their own way.
>
>The more cross breading that takes place on earth because of the increased
>transportation and mobility of people, the more we act as one machine,
>instead of 20 separate machines and the more we become one homogeneous
>race. This is not a surprise, it's exactly what you would expect from
>evolution.

I don't think so. According to the Darwinian doctrine, new species
arise as a result of mutations in a large population. The larger the
population, the more numerous the mutations and the greater the
chances for successful ones. We have more people on earth than ever
before. Where are the new species of intelligent humans? Sure we have
lots of races but a race is just a variety, not a separate species.

PS. This discussion is interesting but, unfortunately, I can only
contribute once in while. I am way too busy at the moment.

Louis Savain

Why Software Is Bad and What We Can Do to Fix It:
http://www.rebelscience.org/Cosas/Reliability.htm
.

Relevant Pages

  • Re: Naturalistic Fallacy
    ... They are a part of Darwinian evolution. ... It is interesting to note that sometimes what we humans prefer to call the ... Few spots remain on earth that are ... ESCAPING each species' needing SOME kind of externality complex if it is to ...
    (sci.bio.evolution)
  • Re: Couple questions for evolution experts
    ... >> process of evolution was occuring, did the human female birth canal ... If it were beneficial for the survival of the species, ... Why is it such a guessing game with humans? ... many atheists agree with those religious persons that evolution and religion are not alternatives. ...
    (talk.origins)
  • Re: Now Steven J claims poverty of the fossil record and that
    ... undergone no evolution at all; you use it to describe entire sequences ... of similar but clearly distinct species each showing microevolutionary ... the present day among living populations), ... specific environment disappears populations revert "back" to its ...
    (talk.origins)
  • Re: Temporal Learning
    ... >>principle about how the adapation works so well in both apes and humans. ... You can believe in the magic of evolution all you ... > species to emerge, there must be a lot of opportunities for mutations. ... It's the environment that defines the species, ...
    (comp.ai.philosophy)
  • Re: Why Should Evolution be Taught as Science in Schools?
    ... how humans came to be and how we fit in the universe"... ... human prehistory and the origin of species does not mean that those ... Macro) evolution has been tested, ... logical that humans genetics would also most closely match primates. ...
    (talk.origins)