From phm@a2e.de  Fri May 25 19:39:45 2001
From: phm@a2e.de (PILCH Hartmut)
Date: Fri, 25 May 2001 20:39:45 +0200 (CEST)
Subject: [Patents] Legal definition of "non-obvious".
In-Reply-To: <20010523011126.A14391@jl1.quim.ucm.es>
Message-ID: <Pine.LNX.4.31.0105252008210.23038-100000@querty.oas.a2e.de>

Ram'on pointed us to

> http://news.findlaw.com/cnn/docs/viagra/viagrapatent.html
...
> As soon as there is something minimally creative or original we have
> something "non-obvious". This is the law.
...
>     62. The question of obviousness has to be assessed through the
>         eyes of the skilled but non-inventive man in the art. This is
>         not a real person.   ...

Maybe it is possible to implement this non-obviousness-testing-robot as a
computer program.  One could then go one step further and define that
the robot be given

- a formalised problem
- a formalised base of prior knowledge

If the robot can come up with the solution that is being applied for, then
the application is deemed obvious.

That could provide an objective standard and would at the same time
eliminate most abstract-logical ideas, depending on how well the program
is written.

Some AI programs are good enough to beat a world chess champion.  It is
hard to imagine that such a program would not be considered prior art if
someone wanted to patent a series of chess moves.  This should mean that
once such inference programs exist, they must be used by patent examiners
to check whether something is really new.  In order to perform this kind
of check, the examiner will have to formalise both the solution and all
the prior knowledge.

Can that realistically be done?  Doesn't it already lead all
abstract-logical patent applications ad absurdum?

For the concrete-physical solutions, the matter is different, because they
are based on new empirical knowledge that can't be obtained by inference
or any other computing operation.

This thought experiment seems to prove from yet another side that the area
of patentability must be limited to problem solutions that are inseparably
bound to matter (containing a new teaching on physical causality).

The German Federal Court already formulated in the 1970s that
patentability must be limited to this area for another reason:  because
anything teaching that is not bound to new concrete-physical structures
will tend to be usable by people without any investment in new equipment,
so that the "need for keeping it freely usable" (Freihaltungsbeduerfnis)
by far outweighs the need for protecting the interests of an inventor.

The other side of this medal is now:  any teaching that is not bound to
new concrete-physical structures will tend to be inventable by an AI robot
without any human inventive effort.

Maybe writing this kind of AI program as a basis for arguing non-novelty
would be very useful.

Of course some patent inflationists will argue that formalising prior art
and problems is in itself an inventive activity.  However that would
require yet another round in patent inflation.  We are not there yet. The
EPO claims that it is offering patents in return for solutions only, and
this has been the traditional doctrine at least in Germany.

--
Hartmut Pilch
Federation for a Free Information Infrastructure        http://www.ffii.org/
Protecting Innovation against Patent Inflation	      http://swpat.ffii.org/
For a software patent free Europe PLEASE SIGN http://petition.eurolinux.org/








