Re: What best evidence for fast AI?

From: Thomas McCabe (
Date: Sat Nov 10 2007 - 16:00:48 MST

On Nov 10, 2007 3:41 PM, J. Andrew Rogers <> wrote:
> On Nov 10, 2007, at 10:41 AM, Алексей Турчин wrote:
> > The only way we can proove that AI is really impossible - is to
> > show that intelligense requiers 'black box' soul.
> Not really. To show that intelligence requires a soul, you would
> have to define 'soul' well enough that we could theoretically
> implement it on conventional hardware. Whether or not it is a 'black
> box' is immaterial; we do not need to know what is going on inside,
> just how it behaves at the interface in relation to the state that is
> measurable.
> J. Andrew Rogers

It would be possible, in theory, to show that intelligence requires
hypercomputation- eg., to build an intelligence, you must also be able
to build a black box with a Turing machine as input and "will
halt"/"will not halt" as outputs. We could still understand the
inputs/outputs, but we wouldn't be able to write an algorithm for it
on any kind of substrate.

 - Tom

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:00 MDT