From: Brian Atkins (brian@posthuman.com)
Date: Tue Jul 02 2002 - 22:15:04 MDT
James Higgins wrote:
>
> At 05:18 PM 7/2/2002 -0400, you wrote:
> >Aaron McBride wrote:
> > >
> > > Please see the posts on "Effective AI Jail":
> > > http://sysopmind.com/archive-sl4/0106/index.html#63
> > >
> > > It's not an exact overlap, but most of this has been discussed
> > > before. Basically, there *might* be ways to trap the AI in the short run,
> > > but once it's smarter than us there's not much hope.
> > >
> >
> >Right, and this added to the rapid increase in hard drive storage
> >capacity currently taking place seems to point towards keeping an
> >in-house copy of the WWW as the safer choice to take. Not only
> >that, but then the AI can access it all using higher bandwidth/lower
> >lantency.
>
> I suspect that bandwidth & latency won't have too much effect until the AI
> has gotten quite smart. Because unless the research is very well funded
> the AI probably won't be able to process the data faster than a typical
> connection could deliver it anyway. Certainly, you would save a few
> seconds here and there due to latency, but a couple of seconds shouldn't
> make that much difference. Besides, this might provide time for the AI to
> ponder.
>
If your AI isn't capable of sucking down a constant OC-12, then that
just means you aren't running it on fast enough hardware :-)
-- Brian Atkins Singularity Institute for Artificial Intelligence http://www.intelligence.org/
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:40 MDT