Re: A call for 1 million programmers

On Dec 6, 9:53 pm, c...@xxxxxxxx (Curt Welch) wrote:
mpfounder <mpfoun...@xxxxxxxxx> wrote:
On Dec 3, 7:47 am, c...@xxxxxxxx (Curt Welch) wrote:
I will insert my comments into your long post.

If you think that was long, check out my posts to comp.ai.pholisophy. That
was one io my short posts on AI. :)

I don't think the solution to AI is a big coding project. I think it's
more like 1000 lines of code tops. I think human behavior is a simple
reinforcement learning algorithm which no one has yet figured out.

I think almost everything Jeff Hawkins writes in his On Intelligence book
is dead on - except for the fact that he doesn't seem to understand the
need for all that temporal hierarchical learning to be there for the
purpose of creating a reinforcement learning system. He seems to have
overlooked the biggest and most important piece of the puzzle in my view.

AI research is not a coding project. It's a research project waiting for
someone to create new insight - to invent something new. I see no purpose
trying to get 10 coders together let alone amillion. You need only a few
smart researchers. At best, one smart researcher might have 10 grad
students working for him implementing his ideas. There's just nothing to
be done by a large set of open-source developers on this. It's just the
wrong approach.

If you haven't tried it, come join the crazy people in comp.ai.philosophy
and try to talk so much as as one guy into believing your approach is the
right one. I'm willing to bet you that you won't find a single guy that
that will agree with your view of what has to be done to solve AI - not
because your view is wrong, but simply because no two people looking at AI
can ever agree on an approach. It's just the way AI works. Everyone
thinks they understand it but no one can agree.

I've been looking at AI for about 25 years now and I've posted thousands of
messages in c.a.p. for the past 5+ years debating just about everything
there is to debate about AI and how to make machines act more like humans.
No one agrees on the solutions or the direction to the solution. Everyone
has a different view about what AI is, what intelligence is, what it means
to be intelligent, what it takes to make a computer act like a human,
whether it's even possible to make computers act like humans, what
consciousness is, whether computers or other machines can be conscious,
whether a machine that acts like a human is conscious or not, whether it's
important whether it's conscious or not, on how long it will take, or how
much it will cost, on whether machines are already intelligent, and on and
on. It's a never ending debate that some of us enjoy even though we spend
half our time calling people idiots and the other half being called an
idiot.

Come to c.a.p. and convince me you have a direction worth working on and
maybe I'll help you. But so far, your approach makes me think your
direction isn't going to go anywhere.

Do you know how the researcher manages to get the 10 grad student working
for him? He pays them, or he first makes a name for himself by publishing
popular papers or books. To get a staff of 10 students trying to implement
your ideas, you have to nearly walk on water. And you think you can get
1000, or amillionimplementing your ideas? You seem to have no clue what
you are up against. Unless you can first prove you have something new, and
something big, you aren't going to get anyone to help you unless you pay
them to work for you.

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

Curt, I've spent the past couple of years overseeing "practical"
software development for my startup. One thing I can tell you is that
it took me about 2 hours to define conceptually in my mind what then
took about 6 man years of programming to implement. I'm a programmer
myself for quite some time, and I know this to be a truth: Software is
tediously slow to produce, even with great programmers. Short cuts
are almost always made, because software details and complexity are
ALWAYS underestimated.

You can't build a good platform for anything with a few programmers,
maybe with 20-30 you can begin to do something. 20 years ago maybe
you could begin to do something with a small core group, but not now.
Case in point: Is an OS conceptually so difficult? No, but to do it
right, with all the bells and whistles, is an enormous undertaking.

With an AI type project that has a chance of making progress there
needs to be enough people working in the same direction, to build a
platform that CAN be agreed on. Everyone having a different view
indeed goes nowhere, and perhaps that is why the AI groups, if you say
correctly, are just a place of disagreement.

Jeff Hawkins is finally starting to gain some small momentum with
Numenta it appears. Perhaps if you feel they are missing on a key
point, then maybe you can add it to their platform. They say that
they have made it quite extensible, and that a good part of it is open
sourced. I myself think they are way to small of a group still,
because there are so many different things to try, tools to build,
ways of integrating what the people using their platform develop...

The point of the Million Programmers is that there are too few big
projects anymore, and no one dreams. Apparently most of the
developers that have responded to this post, are very concerned that
I'm going to make some money on this. I find it amusing, sad and
absurd, all at the same time, but it shows how little they are aware
of how much the VCs, Execs, the rich 2%, extract. A startup typically
gives the hired CEO 8-10% equity, and he is hired to manage the
company. I guess everyone that has responded expects that I'm going
to contribute my ideas, my software for nothing, and work for a
engineering salary or less. I don't give my time away, I do expect
compensation for my ideas, and I protect them.

If everyone could just ignore the fact that they might have to pay
something in eventually, they might see that there is no risk to
this. They can always form their own groups or decline to
participate.
.

Relevant Pages

  • Re: Should we broaden the topicality of this group?
    ... I write virtually all of my work programs in pure ISO C. ... programmers who know what pure ISO C is who use it... ... a detailed answer to a completely platform specific question which would ...
    (comp.lang.c)
  • Re: A C Adventure: your comments are welcome
    ... Most programmers write for Windows. ... language and what parts are specific to the implementation. ... write to it is an error, regardless of platform. ...
    (comp.lang.c)
  • Re: The case insensitive #include statement horror...
    ... programmers follows the programming style guide rules. ... expect the platform or the toolset to help you here. ... old Unix filesystems worked such, ...
    (microsoft.public.vc.language)
  • Re: Generating Gaussian distributed random variable using C++
    ... programmers want their programs to be both safe and as fast as possible. ... Who know the most details about any particular platform? ... The compiler writers for that platform. ... problem-domain knowledge of the library writers. ...
    (comp.dsp)
  • Re: About method comments
    ... When intensively working to solve a problem our programmers only see the ... does it enter their consciousness. ... Apparently at the development stage method comments don't add value, ...
    (comp.lang.smalltalk)