From: Matt Mahoney (matmahoney@yahoo.com)
Date: Sat Sep 27 2008 - 15:50:08 MDT
--- On Fri, 9/26/08, sl4.20.pris@spamgourmet.com <sl4.20.pris@spamgourmet.com> wrote:
> I just read the following:
> http://www.sl4.org/wiki/SoYouWantToBeASeedAIProgrammer
>
> Eliezer, you plan to use Java??? You must be kidding, right?
I agree. We aren't even close to the point of writing code. How likely is a hard takeoff in the form of an alliance of intelligent, evolving internet worms? They could use multiple points of attack:
- Exploiting known security flaws in common software discovered by searching security/hacker websites for exploits. (There are thousands of known exploits, as well as automated tools for discovering them like nmap and virustotal.com).
- Discovering new exploits by analyzing source code and compiled code. (Currently, no security tools can protect against undiscovered attacks).
- Using natural language to trick users into installing the worm. ("Please enter your administrative password to install 79 updates").
Intelligent worms could communicate their successes to other worms, for example, by posting encrypted code to hacked websites. They could reproduce very rapidly, modifying their offspring either randomly or intelligently, thus evolving to forms that evade whatever methods we use to detect and remove them. The challenge for us is how to respond to a global attack using computers that appear to be OK but are likely to already be infected with intelligent agents monitoring everything you do and feeding you false information.
And this is just one scenario. We have to consider all the others too.
-- Matt Mahoney, matmahoney@yahoo.com
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:03 MDT