[FRIAM] AI advance

[Roger Critchlow]

On Tue, Jan 31, 2017 at 11:50 AM, Pamela McCorduck <pamela at well.com> wrote:

>
> > On Jan 31, 2017, at 7:32 AM, Steven A Smith <sasmyth at swcp.com> wrote:
> >
> >
> >> " AlphaGo itself isn't scary it's what comes next and so on and how
> quickly these advances are progressing that give some great minds cause for
> concern."
> >>
> >> I just hope it comes soon.  Humans aren't making very good decisions
> lately.
> > Maybe... but somehow I'm not a lot more confident in the *product* of
> humans who make bad decisions making *better* decisions?
> >
> > Coal Fired Power Plants, Internal Combustion Engines, and even Smart
> Grids make decisions based on their creators values all the time. Why would
> an AI created by short-sited, narrow-minded humans do any better?
>
> For one thing, it can search a larger search space for solutions.
>

The AI's can search a larger space for solutions, but there could be an
even larger space to search which was discarded by faulty assumptions the
builders brought to the problem.  There's room for an inconceivable amount
of space out there in parameter land and many wonderful ways to leave your
thumb on the scales.

-- rec --
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://redfish.com/pipermail/friam_redfish.com/attachments/20170131/96cdc476/attachment-0002.html>