Please log in.
Before you can vote, you need to register. Please log in or create an account.
Business: Information service
Automation of the Inventive Process   (+3, -4)  [vote for, against]
computer based invention algorithms based on human querying

A successful patent application in the United States (i.e. one that matures into a US patent) must undergo several tests to establish it's patentability. These tests can be basically broken down into four categories: usefulness, enablement, novelty and nonobviousness. Recently there have been several proposals to automate the tests performed during the patent examination process.
   US Patent No. 5,721,910 entitled "Relational Database System Containing a Multidimensional Hierarchical Model of Interrelated Subject Categories with Recognition Capabilities", published February 24,1998 and assigned to Exxon Research and Engineering, discloses an automated method of classifying technological publications and abstracts into various business, scientific or technical fields. Classification is very important to the establishment of what constitutes a patentable invention for it defines the scope of the technological art that the invention can be applied. Employing the automated classification system such as disclosed by Patent No. 5,721,910 to a list of technical documents and references may be the first step in establishing a basis for determining whether a patent application claim is novel and may also by applied to establishing non-obviousness.
   US Patent No. 5,774,833 entitled "Method for Syntactic and Semantic Analysis of Patent Text and Drawings", published June 30, 1998 and assigned to Motorola, discloses how to analyze a patent application in accordance with the common format utilized by the US patent office.
   US Patent No. 6,311,176 entitled "Method and Data Structure for the Computer-Aided Management of Developments", published on October 30, 2001 and assigned to IPCentury AG, discloses how to organize patent data in a classification system with low redundancy to minimize the official examination work which establishes the validity of patent applications.
   While these references are directed mainly toward the examination of the patentability of patent applications it is clear that such techniques might easily be applied to the establishments of novel, non-obvious inventions. Typically most novel, non-obvious inventions come about by a combination of several different concepts from differing fields of endeavor wherein there is a recognition of an analogous problem being solved in one field of endeavor which may be applied, with appropriate modification, to a second field of endeavor. By using the techniques disclosed by the above references in creating a first classification system, which classifies publications by the problems being solved, and a second classification system, which classifies publications by technological fields of endeavor, a new invention may by developed by performing an appropriate cross-reference of the two classifications.
   For example an inventor who in the field of endeavor of fluid or gas dispensing may find it advantageous for some reason to adhere a glass nozzle plate to a ceramic substrate, although there may be no teachings in the inventors field of endeavor of a proper manner of connecting the ceramic substrate and glass nozzle plate given the specific parameters and constraints desired by the inventor. However, there may be a teaching of a bonding method between ceramic and silicon which solves the inventors problem exactly but this teaching is in such a far removed field of endeavor and in such a different context that it is imperceptable to the inventor or anyone else skilled in the art. However a solution might easily be found by properly querying, based on the inventor's desired specifications, a classified database structured in terms of problems being solved using methods such as in US Patents 5,721,910 and 6,311,126. This, of course, is a very simple example of the potential of an automated invention strategy and relies a great deal on the proper querying performed by the inventor. In addition any new invention would have to be weighed in terms of it's practicality and potential earning versus cost of production and any automated type of invention algorithm which produces new inventions from a large survey of prior inventions would have to be verified as reliable by human based analysis at least in some part. Fundamental breakthroughs also would not be produced by any conceived automated invention. However the vast majority of patents comprise improvements over established technology rather than landmark breakthroughs. As further developments in searching algorithms and classification database formation continue, better and more reliable invention algorithms may be produced which factor in variables depending on projected earning versus cost of production for the invention and various other desired variables that could minimize the degree of querying of the inventor.
   Problems which may arise in the development of automated invention include a potential legal problem in establishing that the invention is "made by man" which is a requirement covered by Title 35 Section 101 of the United States Code that covers what constitutes patentable subject matter. However the statute does not specify what percentage of contribution to the invention must be made by man and as long as part of the invention is determined by human intervention it would seem that this part of the statute is met.
-- blaise, Sep 08 2002

12864 http://www.halfbakery.com/idea/12864
The Golden Age. Brace yourself. [General Washington, Sep 08 2002]

off the cufflinks http://www.halfbake...f_20the_20cufflinks
Hurry hurry hurry! Your votes are still eligible! [General Washington, Sep 09 2002]

Semantic Web http://www.semanticweb.org
A new type of web that will "understand" what is asked and give intelligable answers [pashute, Sep 30 2002]

http://www.imagination-engines.com/ [blaise, Jul 19 2005]

http://www.automatinginvention.com/ [blaise, Jul 19 2005]

US Patent 6,847,851 http://www.freepate...ne.com/6847851.html
US patent 6,847,851 is a patent on an improved PID controller that has admittedly been created by non-human intelligence (specifically a type of AI called genetic programming). [blaise, Jul 19 2005, last modified May 07 2006]

US 7,904,453 http://www.google.c...m=1&ved=0CCgQ6AEwAA
cites this idea as prior art [xaviergisz, Jul 30 2011]

possibly so.

where were you in the queue when paragraphs were given out?
-- po, Sep 08 2002


This is practically unreadable.
-- waugsqueke, Sep 08 2002


<summary>There exist ways of classifying patent applications by textual analysis; A similar classification of existing inventions might allow for the automatic generation of new inventions.</summary>
-- hippo, Sep 08 2002


Is this what Vernon used to be like? I am sorry that evidence of his existence can only be inferred by newcomers like me. Him and the Great Troll Migration of 2001.
-- General Washington, Sep 08 2002


I think we've been hit by the latest troll virus: automatic postings.
-- FarmerJohn, Sep 08 2002


+1 for sincerity. hi to blaise.
-- po, Sep 08 2002


<summary>There exist ways of classifying patent applications by textual analysis; A similar classification of existing inventions might allow for the automatic generation of new inventions.</summary>

thanks hippo, back to business
-- po, Sep 08 2002


Hey, are you two related?
-- General Washington, Sep 08 2002


Scotch that earlier comment. I have now done some due diligence in Vernon Studies... oof. Anyone who doubts the Halfbakery's higher purpose should take in the 12864 linky.
-- General Washington, Sep 08 2002


My read, fwiw: the proposed invention is a cross-referencing system that bridges two patent-classification methods: those which classify inventions by the problems that the inventions purport to solve and those that classify inventions by their particular field of endeavor.

The balance appears to be a discussion about how a system to automate queries against this cross-reference db might be made and what patentability factors such queries might take into consideration.
-- bristolz, Sep 08 2002


Title's not very descriptive, though.
-- phoenix, Sep 09 2002


There are paragraphs, they've just been included in the manner that I was taught to include them, just leave a slight indent as opposed to a whole line.
-- kaz, Sep 09 2002


those indents were an afterthought.
-- po, Sep 09 2002


No, the indents were included in the original idea but they were formatted incorrectly. That has now been fixed.
-- bristolz, Sep 09 2002


You can't just juggle together a million disparate ideas in your semantic blender and hope what comes out is novel, non-obvious and marketable. Hey- let's apply the principle of gyroscopic stabilisation to the domestic toaster. Brilliant- a $1000 toaster that never falls over! Ok, having made a mint with that one, howabout a clockwork snowboard? Get a million monkeys on the job, why don't you?

It's not called "prior art" for nothing. It's possible to analyse Bach with statistical measurements of notes per second, bars per note, cubic trills per fortnight...but what do you really learn about music? Try generating music with algorithms and see how many fans you get.

I'll lay my cards on the table: I invent electrohydraulic machines for a living. My humble opinion is that the creative process of invention is one of the cornerstones of human intelligence. As such it is not tractable to the brute force and pure ignorance of semantic AI.
-- shameless_self_reference, Sep 09 2002


I'm somewhere in between. The hardest part, I agree with s_s_r, is the making of those new connections, which is highly valued in adult humans ("wit"). But surely you can be helped -- prompted -- led to water. By a machine with no prior meme-connections to ignore. This is what I was getting at with my poorly-received posting 'off the cufflinks' [link].
-- General Washington, Sep 09 2002


What you're suggesting is a bit like Brian Eno's Oblique Strategies... a perturbation, to be sparingly injected into the creative process when it stalls. Sort of like injecting noise into your neurons to shake out the half-baked ideas... Mmm, I feel an idea coming on.......

Problem is, I think the system as proposed by [blaise] would have no *understanding* and therefore an extremely low signal-to-noise ratio. So low you would spend years wading through electrostatic teacups before stumbling upon a better mousetrap. I'm with Dawkins on this one- evolution does not favour a gene which increases mutation!
-- shameless_self_reference, Sep 09 2002


Since even one invention is good, how long do you say would it take for a human to do this and prove the point? IMHO you are simply a bit ahead of time. In the near future (5-7 years) searches on the web (and of course in the patent office) will be "smart searches". Thus nobody will be wading through tons of noise, and if you are looking for a new mouse-trap, you'll easily find your answer in the combined fileds of teacups and kitchen utensils, electrostatics and DSP signal to noise ratio elimination. In any case, take a look at http://www.start2think.com (nothing to do with this annotation).
-- pashute, Sep 30 2002


[Pashute], to make a better mousetrap than the ones currently on sale, an engineer needs more than the words contained within patent libraries. An *understanding* about how a mouse operates and can be made to fail, 3D thinking, knowing how to use production processes and materials and what their limitations are, a gut feel for mechanisms... these can only be gained from many hours in the workshop working on prototypes. They cannot be done by computers crunching words. Do not underestimate how much engineering thinking is non-verbal!
-- shameless_self_reference, Sep 30 2002


That thinking can't be done by computers crunching words but it can be done by AIs crunching concepts.
-- Voice, Jul 31 2011



random, halfbakery