Re: Constructability of AI

Bruce Edmonds (b.edmonds@MMU.AC.UK)
Tue, 22 Jun 1999 15:28:28 +0100


> No, you did not understand my point. I was saying that the
> notion of constructibility is not relevant for describing
> evolving TMs. I believe that these are 2 different statements:
>
> 1. Evolving TMs (that pass the LTTT) are not constructable.
> 2. The notion of constructability is not applicable to evolving TMs.
>
> My impression is that the latter statement is more correct.
> A fly may be wingless but an elephant can not be wingless
> because the notion of wing is not applicable. It looks like
> "non-constructability" is not a quality in the same way as
> being "non-existent" is not a quality.

You are right, I was not very clear in the paper on this.

My intention was merely to motivate the need for essentially exogenous
(i.e. from outside) information by showing that constructable TMs might
not be sufficient.

I was thus wishing to counter the argument: "we can be implemented as
TMs, therefore a TM can pass the TT, and if it is a TM we can construct
it". But I was seeking to do so by avoiding the (sterile and abstract)
argument of whether we could be implemented as TMs (not that I think
that it true, but I don't think it is refutable - a la `Pragmatic
Holism' paper).

Regards.

--------------------------------------------------
Bruce Edmonds,
Centre for Policy Modelling,
Manchester Metropolitan University, Aytoun Bldg.,
Aytoun St., Manchester, M1 3GH. UK.
Tel: +44 161 247 6479 Fax: +44 161 247 6802
http://bruce.edmonds.name