A question for OOP experts. - page 30

 
Vladimir Perervenko:

No one has yet surpassed the Russian vernacular in terms of speed of perception and effectiveness. British scientists have proven.

That's because Russian swear words have soul + clear, scathing instructions.
 

I have been thinking about AI for a few weeks now. It is known that the theoretical basis of this issue is far from clear and systematic, but it is becoming increasingly clear that no one really understands how humans think. The irony is that Consciousness rules the World, but does not know itself. The hidden mechanisms of Mind work as if in the background, maintaining or restructuring a dynamic model of reality, but the "mirror" - the understanding of the self - constantly falls out of it. I have tried to focus on it. To begin with, I defined two key concepts of AI:

  1. Intelligence - conscious mental activity.
  2. Knowledge base - a simplified structured model of reality that acts as an "object" of interaction with Mind.

That is, Reason and its "object", Reality, interact in a continuous process where Reason models Reality, reflecting it as an environment of object interrelations (through "object prism"), and Reality (according to Reason itself), generates and destroys it. We will not discuss a philosophical question "whether Reason is an accidental or regular product of Reality", but we will assume a priori that Reason and Reality are meant to interact.

In the context of our problem - AI is a model of Reason, Knowledge base is a model of Reality, our task is to model their interaction.

This thread raises questions about the applicability of OOP, which I have identified as one of the most important tools for AI implementation due to inheritance, encapsulation and polymorphism, but as I said earlier, I do not consider this tool effective for AI in its current presentation . The "world picture" model is too vast to be described by a standard programming language. That is, the Knowledge Base - our "reality model" - has so many objects, properties and patterns that standard programming with its heavy syntactic "trailer" will inhibit description. Imagine how much time and resources such work would take. In addition, the standard OOP code has an unacceptable flaw - it is static - which means that "training" of AI will turn into endless rewriting the source code of the Knowledge Base. Reality constantly reveals itself to Reason in the process of interaction with it, and it is important that Reason can promptly "rewrite" the model of reality, changing the "source code" of the Knowledge Base on the fly. This reveals a "defect" of the standard OOP-code in relation to AI - it does not support dynamism of the Knowledge base structure and requires its "manual" rewriting. But it is clear that the Knowledge base, as well as the database requires constant updating, as well as restructuring, which in this context is a consequence of AI learning. Therefore, Knowledge Base cannot be a "working object" of Intellect, being structurally described by static OOP code.

Let's move on to neural networks. Can they help in creation of the Knowledge Base? I believe - absolutely, but not completely. NS can represent Reality in the form of abstract images and patterns reflecting forms of objects, states and interactions and this material can be classified and inherited, but I believe the drawback is excessive "interpretability" of represented data in description of clear and unambiguous relationships, formulas, laws.

Therefore, questions of technology and methodology of Knowledge Base construction are open. And we have not yet started to raise the problems of the "engine" - the Intellect.

 
Реter Konow:

I have been thinking about AI for a few weeks now. It is known that the theoretical basis of this issue is far from clear and systematic, but it is becoming increasingly clear that no one really understands how humans think. The irony is that Consciousness rules the World, but does not know itself. The hidden mechanisms of Mind work as if in the background, maintaining or restructuring a dynamic model of reality, but the "mirror" - the understanding of the self - constantly falls out of it. I have tried to focus on it. To begin with, I defined two key concepts of AI:

  1. Intelligence - conscious mental activity.
  2. Knowledge base - a simplified structured model of reality that acts as an "object" of interaction with Mind.

That is, Reason and its "object", Reality, interact in a continuous process where Reason models Reality, reflecting it as an environment of object interrelations (through "object prism"), and Reality (according to Reason itself), generates and destroys it. We will not discuss a philosophical question "whether Reason is an accidental or regular product of Reality", but we will assume a priori that Reason and Reality are meant to interact.

In the context of our problem - AI is a model of Reason, Knowledge base is a model of Reality, our task is to model their interaction.

This thread raises questions about the applicability of OOP, which I have identified as one of the most important tools for AI implementation due to inheritance, encapsulation and polymorphism, but as I said earlier, I do not consider this tool effective for AI in its current presentation . The "world picture" model is too vast to be described by a standard programming language. That is, the Knowledge Base - our "reality model" - has so many objects, properties and patterns that standard programming with its heavy syntactic "trailer" will inhibit description. Imagine how much time and resources such work would take. In addition, the standard OOP code has an unacceptable flaw - it is static - which means that "training" of AI will turn into endless rewriting the source code of the Knowledge Base. Reality constantly reveals itself to Reason in the process of interaction with it, and it is important that Reason can promptly "rewrite" the model of reality, changing the "source code" of the Knowledge Base on the fly. This reveals a "defect" of the standard OOP-code in relation to AI - it does not support dynamism of the Knowledge base structure and requires its "manual" rewriting. But it is clear that the Knowledge base, as well as the database requires constant updating, as well as restructuring, which in this context is a consequence of AI learning. Therefore, Knowledge Base cannot be a "working object" of Intellect, being structurally described by static OOP code.

Let's move on to neural networks. Can they help in creation of the Knowledge Base? I believe - absolutely, but not completely. NS can represent Reality in the form of abstract images and patterns reflecting forms of objects, states and interactions and this material can be classified and inherited, but I believe the drawback is excessive "interpretability" of represented data in description of clear and unambiguous relationships, formulas, laws.

Therefore, questions of technology and methodology of Knowledge Base construction are open. And we have not yet started to raise the problems of the "engine" - the Intellect.

AI is anything that allows one to solve problems without a precise algorithm. That's all it is.

It could be a neural network, a Knowledge Base, or your own invention.

That's the key concept of AI.

The OOP nonsense is, well, as usual.

 
Koldun Zloy:

AI is anything that allows you to solve problems without a precise algorithm. That's all it is.


It's more than that. For example, how do you analyse whether or not you can cross the road in front of incoming traffic without any numerical data? A child will not be able to determine this, but over time, without any instructions, this understanding comes to pass/knowledge.

 
Alexey Viktorov:

Not only that. For example, how do you analyse whether or not you can cross the road in front of incoming traffic without having any numerical data? A child will not be able to determine this, and over time, without any instructions, this understanding comes to pass/knowledge.

May I ask what exactly it is that you don't like about the common definition of AI?

 
Koldun Zloy:

Can I ask what exactly you don't like about the common definition of AI?

What makes you think I don't like it? I thought it was your personal definition and just added my own thoughts.

 

Let's move on to the issue of Intelligence. What does "conscious mental activity" include and on what mechanisms is it based?

The first and main function of Intellect is the work with"meaning constructions", through which the interaction of Reason and Reality is realized. A meaning construction is a product of Reason's "look" at Reality, reflecting its fragment through its subjective (sometimes creative) prism.

The meaningful construction conveys a "snapshot of a fragment" of Reality as a complex of objects and properties in "living" interaction. The assembly of meaning constructions fragmentarily models Reality or "pseudo-Reality", carrying an objective or subjective messagefrom Reason, or to Reason. It is a semantic 'feed' - thought, knowledge, attitude, and more... Semantic constructions are disassembled, assembled or copied by cognitive functions, among which are known logical transitions from cause to effect, "running" through hierarchical branches of categories or associative chains. The incoming semantic construct undergoes stages of disassembly, evaluation, extraction or modelling of context, and decoding of the source's subjective prism. In response to the incoming, Reason generates its assembly using a set of logical and analytical mechanisms and its Knowledge Base, and the "recycled" material can integrate in parts into experience and even rewrite the canons of perception. These are the general principles of Mind's work with meaning, where each stage is a complex mechanism of which we have a superficial understanding.

ZZY. Meaning constructions are a product of Intellect's work and the cast of Reality presented by Reason is NOT a building block of the Knowledge Base, which merely supplies the assembly material of meaning constructions, the object order and inherent inheritance of cognitive functions.

 

These semantic constructions have long been called semantic links.

Artificial Intelligence cannot perform conscious, even mental, or non-conscious activity - there is no one there to be aware of it.

 
Dmitry Fedoseev:

These semantic constructions have long been called semantic links.

Artificial Intelligence cannot perform conscious, even mental, or non-conscious activity - there is no one there to be aware of it.

I would argue. Awareness, a cognitive function that parses semantic constructions and establishes logical links to a subjective "world picture", may well be modelled at some stage of AI development. Actually, from the very beginning, AI development will focus on the implementation of "awareness" - i.e. - a mechanism for disassembling some "semantic relations" in order to "produce" new semantic relations. I stopped seeing awareness as something mystical. It is simply a mechanism.
 

Awareness is the translation from subconsciousness to consciousness.

And consciousness, is:

Созна́ние — состояние психической жизни организма, выражающееся в субъективном переживании событий внешнего мира и тела организма, а также в отчёте об этих событиях и ответной реакции на эти события.

Artificial intelligence is more like a trained animal.