Steve, I believe you meant to sent this to the list(My email client
usually replies both to the list and sender's email address, looks like
this is confusing for most people, I'll just reply to the list from now on).
On 11/1/2015 2:32 PM, Steve Richfield wrote:
Rao.
On Sat, Oct 31, 2015 at 9:16 PM, J Rao <[email protected]
<mailto:[email protected]>> wrote:
Since most of us believe AGI needs to be trained instead of
programmed, the only thing can be shared would be training
data/scenarios, and for these I personally would like to leverage
existing open source data and commercial products intended for
humans as much as possible.
The people talking about training with text mostly want to use
Wikipedia and other sources of information, which would be most easily
accessed with a programmable web crawler as I envision for my own
project. The web crawler needs are IDENTICAL regardless of whether you
are using the data to train, or looking the data over for to find
expressed needs, except the training people have no need to respond to
what they are using for training. Hence, training applications have
fewer demands on the crawler.
Assume we do use text (it's not a given, as you can see some people are
considering games), we may not need to crawl for them, for example you
can just download the entire Wikipedia site as a huge xml file. If we do
need to crawl a site, we can just use one of the open source crawlers
(and you should probably look into them too...)
You may need some supporting software like web crawler if you want
to build a particular application using AGI, but that is only
specific to that application.
That seems obvious to you, but NOT to me. I believe that with a good
interface that provides for suitable guidance of the crawling process,
that substantially ALL AI-related crawling needs can be met with the
SAME software. For example, a good crawler would have a list of
domains to be crawled that would include wildcard capability, and a
list of domains to be avoided, also with wildcard capability.
Perhaps you could provide an example of something that would be
difficult/impossible for a flexible crawler to do?
See above, the problem is not the difficulty, it's the lack of need to
actually implement this ourselves.
Steve
====================
On 10/30/2015 4:00 PM, Steve Richfield wrote:
I noticed a different viewpoint to my discussion regarding my
startup:
ANY major AI/AGI project is going to need pretty much the SAME
support software, regardless of whether it is weak AI or AGI.
Many of the subsystems I was intending on building are the
SAME subsystems anyone else developing other AI/AGI systems
will need. For example:
1. A web crawler that scans the Internet and presents
postings along
with metadata (name, contact information, etc.)
2. An idiom dictionary, since natural language includes SO many
idioms. You won't even be able to talk much with your AI/AGI
without this.
The early BIG money in computers was made in operating systems
rather than in applications software. I suspect the same will
be the case with AI/AGI.
So, is anyone here interested in writing AI/AGI support
software, or:
1. do you intend to use subsystems that other people write?
2. do you intend to write EVERYTHING needed to support AGI
operation?
3. have you even thought about such issues?
Steve
*AGI* | Archives
<https://www.listbox.com/member/archive/303/=now>
<https://www.listbox.com/member/archive/rss/303/26346070-1cd82ca6>
| Modify
<https://www.listbox.com/member/?&;>
Your Subscription [Powered by Listbox]
<http://www.listbox.com>
--
Full employment can be had with the stoke of a pen. Simply institute a
six hour workday. That will easily create enough new jobs to bring
back full employment.
-------------------------------------------
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
