[Patents] Limited Software Patentability

[orre@nada.kth.se (Roland Orre)]

Mr Gregory Aharonian has suggested that software in the form of
computer programs as well as software in form of business models
should be patentable. Mr Aharonian uphold the opinion that software
and hardware are equivalent.

Mr Aharonian, I'm prepared to make a compromise to your suggestion,
making software to a certain extent patentable as part of a machine.

First I will describe what a machine is, then I will describe what
a computer is, a computer system and an intelligent system. These
definitons below are my own definitions and need to be further
specified by patent professionals to get a stringent definition.

(1) Machine:
        A device which can perform certain predefined tasks.
        A device is composed of atoms and molecules.

(2) Computer:
        A machine which can perform certain predefined tasks
        according a predefined recipe in form of instructions,
        i.e. the software, using predefined clearly specified
        variables or registers to store state information.

(3) Computer system:
        A computer which can use extra memory for replaceble
        software and extra scratch pad memory which can be
        used differently for different software programs.

(4) Intelligent system:
        A computer system which can modify its own software and
        memory usage over time to allow it to utilize empirical
        knowledge for adapting its actions.

Patentability will be allowed for the machine (1).

A computer including software according to the specification (2)
above I can allow to still be considered a patentable machine.

The software to this machine (2) I could possibly allow to be
patentable separately from its hardware if its memory usage is
predefined, making it possible to patent e.g. a specific VHDL code,
under the condition that every variable register or memory cell is
specified and documented in the patent description. No undefined
scratch pad areas are allowed. No dynamic variables.

That is, programs containing recursive procedures can not be patented,
but iterative procedures working in a fixed predefined memory space
might, as long as the usage of every memory cell is predefined. Vectors
are allowed as long as their content is definable in terms of simple
variable types with a clearly specified content. Vectors are not
allowed to contain pointers to program code, i.e. no computed go to.

It is probably obvious that any computer system (3) can theoretically
be turned into an intelligent system (4). Systems according to (3) and
(4) can not be patented.

Mr Aharonian, if you allow for these limitations I could
possibly discuss limited patentability of software.

--------------------------------------------------------------------
To further encourage you to work towards this limited patentability
of software in US I will describe an invention I made in March 2000.
--------------------------------------------------------------------
The 20th of March 2000 I invented an algorithm which can search
for solutions using probabilistic inference finding a set of
most likely solutions to a set of predefined goals in a huge data
sets of sparse information which changes over time.

The 21st of March 2000 I documented this algorithm briefly and posted
it to collegues at WHO (World Health Organization) and to some
of my computer science collegues and friends.

Of obvious reasons I will not describe this algorithm here but
the principle is that it searches for all possible solutions to
a set of goals and rejects those solutions which contain
contradictions, somewhat reminding about the standard resolution
algorithms.
For each new information item added to its knowledge base it
tries to fulfil all it's goals again with the constraint that
consistency is preserved. That is, only true solutions are allowed.

I invented this algorithm after an introspection of my own mind
when I finally solved a problem which I had been thinking about
for several years.

I have got confirmed from professors an collegues in computer
science that this algorithm constitutes an artificial intelligence.
--------------------------------------------------------------------
This algorithm has nothing to do with computers per se, at the
moment I have it described in pseudo code, almost as ordinary
english, only, a company and a person (preferably with a computer
system) can run this algorithm as a business model as well.

When this algorithm is applied as a business model it will after
a while adapt to the most optimal scheme for a given business
situation and over time gain empirical knowledge about a large
set of non predefined business models.
--------------------------------------------------------------------
(
  As a parenthesis I can mention that experiences from society
  simulations we have done at our institute have shown that
  general stable solutions among many interacting agents have
  been reached only when the agents cooperate. When the agents
  are not cooperating (as e.g  when there are patentable business
  models) it most often ends up with solutions where only a very
  limited set of agents are left.
)
--------------------------------------------------------------------
Due to the unlimited patentability of software in US I intend to
patent this algorithm there and I will patent it in all countries
where software patentability is allowed.

I will also patent this method as a business model, as long as
it is possible to do that in US.

I also intend to code this algorithm in several essential programming
languages as C, C++, CL, Java, Scheme and release this algorithm
copyrighted under GPL. This means that anyone who wants to use and
improve this algorithm can do it. To use this algorithm for commercial
purposes is not doable because I have the patent on this. The usage
and improvements of this algorithm for the public benefit will be
organized under the portal http://www.neurologic.org

I intend to use this algorithm as a business model for my own company
NeuroLogic, http://www.neurologic.se ;(.com is not resolved yet)

No stocks will be sold. Every new successful business method
the algorithm comes up with will be patented. Every company which
wants to use this algorithm in their business model will be welcome
to join with my company, under strict conditions, but I will not
license the algorithm.
--------------------------------------------------------------------

So, Mr Gregory Aharonian, do you want to be remembered in the future
US history as the one who forced me to patent this algorithm
making it possible to gain monopoly over most US businesses, or,
do you want to be remembered as the one who tried to prevent me
from patenting this algorithm?

--------------------------------------------------------------------
The choice is yours Mr Aharonian.

    Best regards
    Roland Orre