--- On Sat, 12/6/08, Steve Richfield <[EMAIL PROTECTED]> wrote:

>> The internet has about 10^14 to 10^15 bits of knowledge as searchable text. 
>> AGI requires 10^17 to 10^18 bits.
 
> This presumes that there isn't some sort of "agent" at work that filters a 
> particular important type of information, so that even a googol of text 
> wouldn't be any closer. As I keep explaining, that agent is there and working 
> well, to filter the two things that I keep mentioning. Hence, you are WRONG 
> here.

The idea of distributed indexing is to give Google (the agent that now filters 
the internet for you) some competition.

>> If we assume that the internet doubles every 1.5 to 2 years with Moore's 
>> Law, then we should have enough knowledge in 15-20 years.

> Unfortunately, I won't double my own postings, and few others will double 
> their own output. Sure, there will be some additional enlargement of the 
> Internet, but its growth is linear once past its introduction, which we are, 
> and short of exponential growth of population, which is on a scale of a 
> century or so. In short, Moore's law simply doesn't apply here, any more than 
> 9 women can make a baby in a month.

The extra information will come from surveillance. If you wanted privacy, you 
would be sending me encrypted private email instead of posting to a public 
forum through a free email service like gmail or yahoo that records all your 
mail to train its AI database and learn your preferences so it can deliver 
targeted ads. You would also use cash instead of credit cards, so that banks 
and stores would not have a record of everything you buy. Publishing 
information about yourself makes your life more convenient. Once we have the 
computing power to index speech and video like we do for text, everything you 
say or do will be public knowledge, because that's what you want.

>> However, much of this new knowledge is video, so we also need to solve 
>> vision and speech along with language.

> Which of course has been stymied by the lack of metadata - my point all along.

No, it is stymied by lack of computing power and training data. You can 
understand what you see because your brain was trained on decades worth of high 
resolution video.

>> Or another way to answer your question, AGI is a lot of dumb specialists 
>> plus an infrastructure to route messages to the right experts.
 
> I suspect that your definition here is unique. Perhaps other on this forum 
> would like to proclaim which of us is right/wrong. I thought that the 
> definition more or less included an intelligent computer.

Experts can be human or computer. When you don't care which, that's AGI.

> You seem to be making the SAME newbie AI error that others here keep accusing 
> me of making, namely, of making over-broad claims and extrapolations for a 
> basically sound core concept. You have an easily testable concept that could 
> easily be implemented on a USENET group with clever client software, or 
> alternatively and with less ability to scale, on a web site with the postings 
> all tucked away in a giant knowledge base. In the beginning, you could 
> announce a very limited domain and crunch the data with string searches just 
> to test the user interfaces, etc.

I don't expect to build a $1 quadrillion system. Anything I could actually 
build (like a text compression benchmark) would probably not be very impressive.

-- Matt Mahoney, [EMAIL PROTECTED]



-------------------------------------------
agi
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/
Modify Your Subscription: 
https://www.listbox.com/member/?member_id=8660244&id_secret=120640061-aded06
Powered by Listbox: http://www.listbox.com

Reply via email to