I'm thinking of something like this: when a person is curious or has a
question about something they can go out in the world - to other persons,
the internet, a book, database, etc. - and acquire new information, which
would then answer their question or questions or lead to additional
questions or thoughts or other answers. This could be part of a process of
problem-solving or making a decision (with both decisions and problems being
questions or perhaps sets of questions). So wondering if computers could be
developed to simulate this process, i.e. the natural process of how the
human mind works when it plans, etc. (as described, for instance, by David
Allen, in "Getting Things Done, chapter 3 on the natural planning process).
And could computers have more choices than the two they now have in
answering every question, i.e. operate on something other than binary?
Randall
----- Original Message -----
From: "Eric S. Sande" <[EMAIL PROTECTED]>
To: <[email protected]>
Sent: Friday, September 28, 2007 9:11 PM
Subject: Re: [CGUYS] Rename the List?
>Since human thinking can essentially mimic that of computers, albeit at
>far slower speeds, I'm now wondering whether and how computers can be
>designed to more closely mimic how humans learn. For anyone who might
>know, have they got very far with this in AI?
Randy, computers can't think as we do as human beings. Not
yet and maybe never, but never say never...
Computer programs and instrumentalites can be designed to adapt
to circumstances within rule sets. If we want say a terrain following
flying device that can go to waypoints and choose action based on
a set of decision rules, we can make it.
But the cruise missile doesn't have free will. It can't decide that the
action is desirable, or (with apologies to the Reverend) ethical.
There is pretty good AI out there in games, but it is all rule based.
Computers at this point literally CAN'T "think outside the box".
But within the "box", depending on how we define it, they can blow
the doors off human beings with regard to what we consider genius.
They do it, in chess for example, by brute force with deep analysis
of probable outcomes. Chess is simple, though, compared to
say deciding whether to program a computer to PLAY chess.
No computer has yet passed the Turing test. Google this and you
will have a better answer than I can give you.
If we create self-aware (conscious) computers, and I'm not sure we
should, what are the possible outcomes?
1) The HAL 9000 scenario. The computer not only is smart and
passes the Turing test, but also is emotional, learns fear and becomes
insane. Bad news if you are depending on it to run your infrastructure.
2) The Colossus/Terminator scenario. The computer is smart but
completely amoral. But it wants to preserve itself at all cost. You
lose.
3) The "Mike" scenario (_The Moon is a Harsh Mistress_, Heinlein).
The computer is self aware and has a sense of humor, sort of. Luckily
it dies before it figures out that it's being manipulated.
Don't know if this was a good answer, but I hope food for thought.
************************************************************************
* ==> QUICK LIST-COMMAND REFERENCE - Put the following commands in <==
* ==> the body of an email & send 'em to: [EMAIL PROTECTED] <==
* Join the list: SUBSCRIBE COMPUTERGUYS-L Your Name
* Too much mail? Try Daily Digests command: SET COMPUTERGUYS-L DIGEST
* Tired of the List? Unsubscribe command: SIGNOFF COMPUTERGUYS-L
* New address? From OLD address send: CHANGE COMPUTERGUYS-L YourNewAddress
* Need more help? Send mail to: [EMAIL PROTECTED]
************************************************************************
* List archive at www.mail-archive.com/[email protected]/
* RSS at www.mail-archive.com/[email protected]/maillist.xml
* Messages bearing the header "X-No-Archive: yes" will not be archived
************************************************************************
--
No virus found in this incoming message.
Checked by AVG Free Edition. Version: 7.5.488 / Virus Database:
269.13.33/1034 - Release Date: 9/27/2007 5:00 PM
************************************************************************
* ==> QUICK LIST-COMMAND REFERENCE - Put the following commands in <==
* ==> the body of an email & send 'em to: [EMAIL PROTECTED] <==
* Join the list: SUBSCRIBE COMPUTERGUYS-L Your Name
* Too much mail? Try Daily Digests command: SET COMPUTERGUYS-L DIGEST
* Tired of the List? Unsubscribe command: SIGNOFF COMPUTERGUYS-L
* New address? From OLD address send: CHANGE COMPUTERGUYS-L YourNewAddress
* Need more help? Send mail to: [EMAIL PROTECTED]
************************************************************************
* List archive at www.mail-archive.com/[email protected]/
* RSS at www.mail-archive.com/[email protected]/maillist.xml
* Messages bearing the header "X-No-Archive: yes" will not be archived
************************************************************************
