RE: webnet & HAL9000
- Posted by Kat <gertie at PELL.NET> Feb 14, 2002
- 488 views
On 14 Feb 2002, at 20:09, C. K. Lester wrote: > > > > Until our scientists (or Kats) start to understand the > > > requisites, they will always be on the wrong path to AI. > > > > How do you define "requisites"...? > > The "requisites" are the most basic elements required to sustain an > intelligence. It's just like in programming where you have to break down > every task to it's most basic elements. > > I can't say: > > PlayEUQuake() > > I have to break it down... to an extreme! This must also be done when > considering a man-made intelligence. What are the most basic needs? But *humans* shouldn't be doing that. Humans would have a biased opinion, and be error-prone int he extreme. Witness again the 600 man-years on Cyc, and Lenat failed *again*. If the human doesn't pre-munge it, then the Ai must,, and if it cannot incorporate it while it's running, then it's rather useless, isn't it? > > and what do you see they are? > > I've got my ideas, but I'll sit down and consider them more fully > tonight. > > We must answer, "The most basic requisite for a man-made intelligence > is..." > > > >