The development of WAGI, Web Artificial General Intelligence, can involve, for example, an intelligence algorithm with two metasystem transitions, as I explained in my previous article “Bloom’s Beehive: intelligence is an algorithm”. In his book “Creating Internet Intelligence”, Ben Goertzel also implicitly describes it. Steps 3 and 6 that I mentioned in my previous article are the most important steps because they identify differences, correspondences, and spatiotemporal relationships are mapped as patterns. A pattern P is said to be based on a mind, when the mind contains a number of specific entities in which P is in fact a pattern. From correspondences, shared meanings, and grounded patterns, abstractions and simplification rules can be derived, while differences prompt evaluation toward possible modification.

For abstraction and simplification processes in which patterns are derived from numerous data events, Artificial Intelligence programs exist and are developed, but they are often dedicated to a very specific niche. When dealing with numerical data, such as in stock market analysis, trading activity analysis, scientific experimental data, etc. or space-time data, like traffic systems, or data based on rules and patterns, like in games, these programs work quite well for their specific niche. What Goertzel is trying in the OpenCog software and the Novamente project is bringing these features to the world of Artificial General Intelligence (AGI). Here, data mining, which involves a great deal of analysis of a linguistic and semantic nature, is of quite a different order. Although there are quite a few programs (eg DOGMA; OBO, OWL: Web Ontology Language, etc.) and a lot of work has been done in the field of ontology (ontology in the field of AI is a “formal and explicit specification of a shared conceptualization”) there is still room for improvement of the rules and schemes that help establish ontologies.

It is here that the daily work of patent attorneys and patent examiners can contribute ideas for development in the field of Ontology. In fact, much of the work of patent attorneys and patent examiners involves the establishment of ontologies. When a patent attorney writes a claim for an invention, which is a specific entity, he tries to conceptualize how the invention can be described in the most general way, while maintaining all the essential characteristics to define the invention. When writing a request, he must take into account all possible components of an ontology that is commonly known as individuals, classes, attributes, relationships, function terms, constraints, rules, axioms, and events, as illustrated below:

• The specific entities on which a pattern is based, of which at least one must be described in detail and which can be claimed in dependent claims, can be considered as “individuals”, the basic objects.

• The claim dependency structure, the so-called claim tree, has various kinds of intermediate generalizations before reaching individual specific entities and can be seen as providing the “classes”.

• Essentially, there is a claim to a list of characteristics that qualify as “attributes.”

• Through the dependency in the claims tree the “relationships” are provided.

• So-called “functional characteristics” that encompass a number of specific entities provide the “function terms”.

• Disclaimers, conditions qualify as “restrictions.”

• If… then the “rules” result in claims dependent on particular combinations of conditional requirements

• The provision of “axioms” is most often done in the description; it amounts to giving a plausible explanation of why the structural and functional characteristics give rise to the described technical effect that the invention has on the state of the art.

• Changes in attributes or relationships that lead to the making of different independent claims are called “events.”

In a clever way, patent attorneys are extremely proficient in this process. With a minimum of features and functional relationships between those features, to ensure as broad a claim as possible without violating the teachings of the prior art, they come to give an ontological definition of an invention.

The entire process of drafting a patent application and, in particular, a successful claim tree depends on the patent attorney’s ability to identify classes and subclasses: hypernyms and hyponyms. In the feature description you will need to use holonyms and meronyms. And in the ideal situation, the broader independent claim has become so generalized that it is prima facie difficult to see what particular types of inventions fall under the conceptualization.

And the differences do not stop there in terms of the state of the art suitable for evaluation towards possible modifications and/or additional industrial applications.

When a patent examiner has to evaluate a patent application, he has to go through this process in reverse order. He has to figure out what specific entities have allowed generalization and he has to imagine what kinds of existing inventions might fall within the scope of the generalized claims. She has to identify which features (structural and/or functional) are responsible for the technical effect on the prior art. From those insights, she can build a search strategy to identify relevant prior art, which anticipates and falls within the scope of the subject matter of the claims. For this search strategy to be complete, it must combine a set of search terms that reflect all the individual essential features that describe the invention. The search will start with some concrete examples of individual entities and synonyms at a level, but when simple search strategies fail, you will need to define (to the extent not already done by the patent attorney) hypernyms and hyponyms of the features and combine these. Either he will have to describe a feature as a set of meronyms or, conversely, a set of features as a holonym. Nasty problems often occur with acronyms that have more than one meaning, i.e. they are holonymous, leading to search results that have too many documents. Then the boolean NOT operator must be added in an additional search statement to filter out irrelevant documents, so-called noise. Close antonyms to negative terms like “no” or “no” can also generate positive results. If the result sets contain too many members, it should be narrowed down, adding more search terms or more specific search terms. Furthermore, search terms that have a defined relationship can be combined in a specific way to ensure a proximity between the terms: this is done with so-called proximity operators, which are more powerful in such cases than simple boolean “AND” operators. “. Conversely, if a result set has very few members, it can be expanded by using more general terms, fewer search statements, or less strict proximities.

In fact, when building a search strategy, the search examiner is doing a very detailed Partial Ontology, and it is a pity (but a logical consequence of the secrecy requirement) that these ontologies are not stored in a publicly accessible database. in analogy with Semantics. website. Furthermore, the patent examining community has created and continues to create a very detailed classification scheme, such as the IPC, which can be appropriately used as inspiration in the development of ontological classification schemes. It would also be helpful to everyone (not just patent professionals, scientists, inventors, and AI ontologists) if search engines like Google and Yahoo finally made proximity operators available. There is much criticism from the world of scientists and inventors about the inadequate results that web-based search engines deliver (see Grivell, L. in EMBO Reports (2006) 7, pp.10-13). The search engines used by patent offices are far superior in many ways. Unfortunately for you, they are not accessible to the public. Either way, AIbot-based crawlers and spiders do not enter deep web databases, where extremely relevant information may be waiting for you.

Ontologies stored in a specific database with links to other deep web databases that are fully searchable in combination with non-crawler and non-crawler data mining bots can be a huge step forward in providing information.

The proximity relationship is a concept that may require more attention in the field of ontology, since it is an indicator of how certain terms are semantically connected to each other. For example, it would be useful to map each term defined in a semantic web to find out the average distance between all documents on the web from each other. Perhaps from such data mining it would result that certain terms have very close average proximities, where both terms have not been defined in the semantic web to have any relation to each other. It would provide a greater degree of ontological mapping. At a more concrete level involving geographic data, such processes are already underway (eg, Arpinar et al. in “Geographic Information Science Handbook”: Developing Geospatial Ontologies and Semantic Analysis).

Ontologies are somehow needed to build a WAGI based Webmind and it is high time that AI developers at Google, Yahoo etc. start working on these issues and prevent Hegel’s OWLs only fly at dusk.