Dear Chris and All,
> "Rust, Chris" <[log in to unmask]>
> Subject: Re: Artificial Designer (was Automaton)
> I feel that this is the core of our debate. John accepted half of my =
> requirement for a designing artefact (solving the problem)
> but, since he chose to ignore what I feel is the most important =
> requirement (identifying the problem, or as Monica puts it,
> identifying a rich version of the problem),
Chris, thanks for clarifying this. Now I understand you meant 'identify the problem _scope_', I agree with that too. I had, however, taken this for granted. (When you said 'identifying the problem', I had envisioned some philanthropic robot sorting out my problems before I even realised I had them! Hence my puzzlement.)
> I don't feel comfortable with bringing in contemporary
> debates about AI because they bring a lot of baggage that
> quickly takes us out of our area of competence as designers.
But surely we have to go there? If design requires intelligence, then artificial design needs some sort of artificial intelligence, doesn't it? To disallow AI and then say we can't ever achieve artificial designing seems disingenuous. I too am no expert on AI, but I am interested in what light it can shed, and indeed what knowledge of the design process can contribute to the field of AI in return.
> what's our Turing test?
Here you put with considerably more elegance, the sole point I was interested in in my original post.
I did use the word 'continuous' with reference to a scale of design problems in a later post; this was careless of me, and not what I meant to imply in my original post either. As we introduce new design requirements the nature of problems change. I do recognise, therefore, that the scale is discontinuous. But some problems *are* more difficult than others. The scale may well be of categories of design problems, but surely they must, in some sense, be an ordinal set of categories, and somewhere there must be a problem (representative of a problem type) that, if an artificial system should produce an acceptable solution, then we can say, 'yes, it has designed'. Janet McDonnell's contribution was most helpful to me in offering two (of, as she points out, many) characteristics of such a problem; 'no definitive formulation' and 'no stopping rule'. (I am less happy to describe them as 'show-stoppers'; human endeavour has overcome show-stoppers before.)
What are the other characteristics?
Regards,
John
John Shackleton
Brunel Design
School of Engineering and Design
Brunel University
Uxbridge, Middlesex
UB8 3PH
UK
01895 266322
|