I believe I can write a simple AGI program in a year.  It would not
convince the worse skeptics but I would hope to show programmers and
enthusiasts that it can:

1. Learn the basics of a human language or a primitive version of one.

2. Learn simple things from discussion.

3. Learn to make (simple non-mathematical) correlations (between 'objects'
of discussion) and generalize based on what it learned through language.
4. Learn the limitations on generalization and on the use of correlations
as objectives.
5. Use reason-based-reasoning.

However, it will not be perfect, and it will become overwhelmed by the
complexity of acquired knowledge.



Now let's say that 5 months go by and I haven't started the program.  Well,
if I am still reasonably healthy and have the same amount of free time that
I have now that would indicate that my ideas probably weren't that great.
Does that prove that my ideas are wrong?  No, but it would indicate that I
do not have every concept that I need to actually start working on the
program.  In other words, it would stand as evidence that there is
something important that I haven't figured out.



Now let's say that I did get the basic program together but it doesn't do
anything intelligent.  That would be a strong indication that my ideas did
not work, that I was missing something.



Let's say that I felt that it was working but no one in this group (for
example) agreed with me.  Then I would have to improve on it in order to
convince a few people.  I don't feel that I would have to convince everyone
but I would have to be able to convince a few people that my program was
actually working - at least to the extent that I am describing in this
message.



Now let's say that a few people thought it was working as far as it went,
but most people simply did not accept that it was a working AGI program.  That
would be a difficult situation but the way I could validate my sense that
it was working (to the limited extent that I described above) would be to
modify it to show that it could learn using with other modalities (using
other forms of IO) and show that I was able to make actual improvements on
the different versions.  At some point a few of the skeptics would start to
recognize that it was essentially learning new things in the way I have
described.



Jim Bromer



-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com

Reply via email to