On Fri, Oct 19, 2012 at 7:09 PM, David Clark <[email protected]> wrote:
> It is quite funny that you would be the first to post on my design.  You have 
> posted some of the best ideas IMHO about the scale and implementation of AGI 
> designs for many years on this site.

Thanks!

> IDE in a browser: The main reason is that you would be able to change/design 
> a system from any browser without any installation or local directories of 
> source, test data, documentation etc.  Everything would be contained in one 
> place.  Backed up in one place.  This facilitates multi-programmers working 
> on the same system, you working directly on many client systems and the 
> ability for programs inside the system to have as much access to these things 
> as a programmer has.  A full featured CMS is built in so that the IDE can be 
> enhanced quite easily.

The modern way of doing these things is to break up the problem and
use a general purpose version control system together with local tools
that operate on text files.

I'm not going to argue that this is a better way of doing things than
the monolithic solution - I happen to believe it is, but that's in
large part a matter of opinion, so argument isn't likely to reach
useful conclusions.

I'll just point out some relevant historical facts: Smalltalk tried
the monolithic way, and is now largely dead and buried; and Smalltalk
had in its day orders of magnitude more resources, both financial and
political, then will be available to your project, and in its day the
conventions were less strongly entrenched than they are now. Based on
these facts I predict a monolithic solution introduced now is unlikely
to be used.

> If a programmer wants to use all his/her memory or just screw things up, it 
> is their system after all!

To be sure. In practice of course it's typically the case that I don't
want to do that, but what I do want to do is write meta-code that
generates and tests code that in some cases could run out of resources
or screw things up. How does your system handle this?

> The system is not open source but that doesn't mean I wouldn't be open to 
> providing some or all the source to selected others depending on the 
> arrangement.  I have plans to use this system for commercial use BUT I would 
> be interested in different proposals that would enhance the path to an AGI.  
> I didn't make this system to squeeze money out of AGI researchers!  I have 
> plans to publish enough source so that any independent code could coordinate 
> with a copy of this system as if it was one.

Open source is something of a binary thing: if you aren't willing to
publish the full system under one of the recognized open source
licenses, you may as well not supply any of it to anyone.

As to whether to publish under an open-source license, it's worth
bearing in mind that the market price of programming tools these days
is $0, so unless you have a business model in mind that I haven't
thought of, there is no possible gain from keeping it proprietary.
Better in my opinion to go open source and aim to make money from
services, consulting etc., though of course this is a judgment call
only you can make.

> Search and Inference: I understand both words but I am not sure exactly what 
> you mean.  My system includes indexes, lists, tables etc.  One benchmark I 
> recently did was to create a B+Tree index on a list with 1 million records of 
> 100 byte length where the first 10 characters were random numbers.  On one 
> CPU on my Quad-Core, Windows PC, with no optimized comparer, I created the 
> index in under 1.5 seconds.

Right, sorry, I don't mean search in the sense of databases that can
be stored on a computer but in the sense of NP-hard problems whose
search spaces commonly exceed the number of atoms in the visible
universe, e.g.

http://en.wikipedia.org/wiki/Boolean_satisfiability_problem
http://en.wikipedia.org/wiki/Automated_theorem_proving
http://en.wikipedia.org/wiki/Machine_learning


-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968
Powered by Listbox: http://www.listbox.com

Reply via email to