So we’re talking about text analysis?  (That didn’t hurt, did it  ? )

Give one example of the kind of productive text analysis you (or anyone else) 
mean[s] – and you’ll find it is impossible and save yourself years of life. And 
you could at least start a productive discussion here. [Note that Steve was 
just specific about his proposed project – and that produced a useful 
discussion].

Lots of people seem to have fantasies about a supposed AGI program that is 
going to become wise and ultimately rule the world through analysing the texts 
on the net. It’s total cobblers. As I’ve pointed out, there isn’t a program 
that can productively analyse the possible combinations of two or three words, 
let alone two sentences, let alone the contents of one or two texts. 

The fantasies are all Chinese room fantasies about how a manipulator of 
meaningless words enclosed in a black box can become supremely wise about the 
outside world, without ever venturing outside.  Fantasies of real world wisdom 
without real world experience.

That’s how science became so relatively wise about the world, right? – by 
scientists staying inside their studies and playing with words and logic? Or 
did Francis Bacon first have to smash that fantasy ?



From: Jim Bromer 
Sent: Tuesday, April 16, 2013 2:08 PM
To: AGI 
Subject: Re: [agi] Re: Summary of My Current Theory For an AGI Program.

On Mon, Apr 15, 2013 at 11:29 AM, Alan Grimes <[email protected]> wrote:

  Mike Tintner wrote: 
    What’s your O.D. ? What’s the end-product of your program? Drawings? 
Buildings? Text-readings? Wtf is it going to DO? Or is that too difficult for 
you to say?

  ... I'm getting sick of these jags you go off on. Last week it was "Well your 
AI doesn't implement true creativity; prove that it does!"

  This week you are ignoring the G in general AI. The word GENERAL in AI, like 
in computer science at large, means "Virtually any" So it must be capable of 
dealing with virtually any problem in virtually any domain using virtually any 
method. So therefore it must be able to learn any abstraction less than equal 
some reasonable complexity metric and it must have the computational 
capabilities to optimize and apply those abstractions. 
  ...



Alan,
My text-based AGi program would be a limited kind of AGI program but it would 
be a proof-of-concept thing.  If it worked then it would be general enough to 
convert it for different kinds of IO actions.  A program that could do some 
genuine learning and derive abstractions from text would be flexible enough to 
modify for conversion to image AGI and so on.
Jim Bromer

      AGI | Archives  | Modify Your Subscription   



-------------------------------------------
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

<<wlEmoticon-smile[1].png>>

Reply via email to