XPost: rec.arts.sf.written, rec.arts.sf.science
From: jclarkeusenet@cox.net
In article ,
droleary@8usenet2013.subsume.com says...
>
> In article ,
> Greg Goss wrote:
>
> > "Michael F. Stemper" wrote:
> >
> > >On 01/23/2014 06:17 PM, Your Name wrote:
> >
> > >> There was an article in the newspaper not long ago about the experts
> > >> working with robots etc. saying that getting to a point of having
> > >> actual computer intelligence is stil decades away.
> > >
> > >It always has been. Kind of like fusion.
> >
> > Well, a lot of the skills that were defined as part of "AI" get
> > redefined out once they're met.
> >
> > Expert Systems?
> > Basic language translation?
> > Grandmaster-level chess?
> > Driving a car on public roads?
> >
> > Nope. Those aren't AI after all.
>
> They'd be AI if they were solved with AI. Instead, they are solved by
> clever humans using computers as tools. Things like Watson and
> self-driving cars are, as currently implemented, cheats.
>
> As I have posted in the AI newsgroups, you can easily set up a standard
> for an AI driver. Run this experiment: allow a person to drive a car
> remotely using only a video camera, measuring their performance as you
> reduce the resolution of said camera. Then you pick a "minimum
> acceptable" point for their ability to navigate the environment. That's
> your Turing Test for AI drivers, getting around with only a crappy web
> cam as your input.
And what happens when your human driver with limited video gets t-boned
by the truck he couldn't see due to the limited field of view of the
camera?
Sorry, but that's not a "turing test for AI drivers", that's an accident
looking for a place to happen.
> Compare that to how Google and others are solving the problem and it is
> easy to state unequivocally that AI is *not* being used.
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
