By: dmcq (dmcq.delete@this.fano.co.uk),
Room: Moderated Discussions
Foo_ (foo.delete@this.nomail.com) on November 1, 2017 4:45 am wrote:
> dmcq (dmcq.delete@this.fano.co.uk) on November 1, 2017 4:29 am wrote:
> > Foo_ (foo.delete@this.nomail.com) on November 1, 2017 4:09 am wrote:
> > > dmcq (dmcq.delete@this.fano.co.uk) on October 31, 2017 6:02 pm wrote:
> > > >
> > > > What kind of intellectual problem do you think is not amenable to being tackled by
> > > > something that assigns a percentage chance of victory to positions given to it?
> > >
> > > Uh? Most kinds of intellectual problems are not, unless the only kind of "intellectual"
> > > problems you know about are the problems we are currently trying to solve with AI.
> > >
> > > Go to a library, in the essays section, and see which books there could have been
> > > written by "assigning a percentage chance of victory to positions to it". Say you
> > > want to write a history of the Middle Ages in Western Europe: how does that work?
> > >
> >
> > Well they're not top class historians yet but using this type of technology
> > to write news stories has been around for a few years now, see for example
> >
> > And the Pulitzer goes to… a computer
>
> Yes... It's merely *presenting* data that's been fed into it (for example, it creates "data-based
> stock reports for AP"). A historian's job, though, is to find data sources, analyze and
> criticize them, then choose a consistent and articulate way to depict history from them,
> so as to make it intelligible. Writing the final form is certainly important (especially
> for books meant for the greater public), but it's far from the main part of it.
>
> Besides, it's not obvious that this news-writing system "assigns
> a percentage chance of victory to positions given to it".
See how well you do at this quiz
Did a Human or a Computer Write This?
Assigning a percentage chance to presented positions is the essence of what neural nets do. The various back propagation and normalization and other procedures applied in improving that is a major object of study.
> dmcq (dmcq.delete@this.fano.co.uk) on November 1, 2017 4:29 am wrote:
> > Foo_ (foo.delete@this.nomail.com) on November 1, 2017 4:09 am wrote:
> > > dmcq (dmcq.delete@this.fano.co.uk) on October 31, 2017 6:02 pm wrote:
> > > >
> > > > What kind of intellectual problem do you think is not amenable to being tackled by
> > > > something that assigns a percentage chance of victory to positions given to it?
> > >
> > > Uh? Most kinds of intellectual problems are not, unless the only kind of "intellectual"
> > > problems you know about are the problems we are currently trying to solve with AI.
> > >
> > > Go to a library, in the essays section, and see which books there could have been
> > > written by "assigning a percentage chance of victory to positions to it". Say you
> > > want to write a history of the Middle Ages in Western Europe: how does that work?
> > >
> >
> > Well they're not top class historians yet but using this type of technology
> > to write news stories has been around for a few years now, see for example
> >
> > And the Pulitzer goes to… a computer
>
> Yes... It's merely *presenting* data that's been fed into it (for example, it creates "data-based
> stock reports for AP"). A historian's job, though, is to find data sources, analyze and
> criticize them, then choose a consistent and articulate way to depict history from them,
> so as to make it intelligible. Writing the final form is certainly important (especially
> for books meant for the greater public), but it's far from the main part of it.
>
> Besides, it's not obvious that this news-writing system "assigns
> a percentage chance of victory to positions given to it".
See how well you do at this quiz
Did a Human or a Computer Write This?
Assigning a percentage chance to presented positions is the essence of what neural nets do. The various back propagation and normalization and other procedures applied in improving that is a major object of study.