Re: purpose of AIs

Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Mon, 13 Dec 1999 12:26:36 -0800 (PST)

Ken Clements writes:

> We also noted that if a program managed to infect the collection of routers on
> the Internet, it could arrange to make traffic patterns look like the net needed
> more routers. This one we *could* see coming because pretty soon the majority
> of our GNP would be going to building routers.

We're going to have purely optical switching (which btw might require a redesign of the lowest transport layer protocols, TCP/IP needs a lot of smarts which are expensive with optical logics) before very long. This is going to bring the costs down.

As to an spontaneously emerged (or deliberately seeded) AI in the global network, it is going to generate a lot of strange traffic between the infected nodes. People are going to notice, and start with countermeasures. The AI will necessarily need to acquire a much lower profile to survive, and hence be slower/dumber, at least for one or two generations of hardware. You need molecular switches to make something truly smart and still desktop-sized.