What kind of power does a “strong AI” need?
In principle, any Intellect of any Object or System is determined by the Behavior of this Object or this System, which, in turn, is determined in the general case by two technologies: PHYSIOLOGY (Mind), as reflex responses to input, and MIND (Consciousness), as meaningful decisions on the assessment of the situation.
The level of the well-known "perceptron" technologies (and those of various "neuro-quantum-semantho-statistical-mathematical-network crap") is just a technology of physiology. That is, the signal came to the input, which means that it performed the corresponding action. But these are just “brainless automata” and now they are called “weak AI”.
However, the natural human intellect is based mainly on the second technology - on the Mind. And this is already a "strong AI." That is what their fundamental difference is in Consciousness.
Technology of Mind, like any other technology, contains a Methodology, as a set of Methods, and Toolkit, as a set of Tools.
So, the main method of Reason is THINKING, and the main tool is LANGUAGE.
From this it follows that if we want to repeat the UI of a person in its computer version, like AI, then all we need is nothing - to implement “computer thinking” along with the corresponding software language.
')
If with the latter, like, no difficulties arise, then, here, with the “artificial-thinking iron”, serious problems still arise. It turns out that for the full implementation of the "machine mind", it is necessary for the computer to be able to understand the meaning of the context at the human level, and still be able to operate with a multitude of knowledge.
And, if the software developers were able to saddle the data well, then with knowledge the situation is not important at all. Today there is not even a standard for the format of their storage and processing. That is, the Database is, and the Knowledge Base is not. And what you call the BZ, do not believe in your ears or your eyes. Above the relational database, we practically did not rise.
Worsening the situation and the fact that linguists with cybernetics over the past half century have not been able to come to a common opinion and approach in teaching computers to understand natural languages and extract meaning from the texts written in these languages. And without this, there can be no progress in creating a "strong AI." So, pundits from many academic circles who are involved in this issue have retreated from this very heavy task that has turned out for them.
From here and results, more precisely, their full absence. Therefore, the vast majority of researchers went into the "weak AI", leaving the problem of creating an "Artificial Mind" to their descendants.
True, there are quite a few who stormed this unaffected fortress with great stubbornness. And there are already obvious successes. The world's first Technology Heuristic-Content Processing of Texts (TESL) has been created, called “Ontology”. It is a pioneering program in its class and in its own way, capable of understanding the meaning of the context and even puzzling with questions. That is, glimpses of the first artificial thinking have already appeared. So not everything is so bad on the fabled fronts. At the very least, we can already report on “MEDIUM AI”.
The dialogue system created on its basis has the name "DUMBOT" (from THINKING ROBOT). Today, different developers started to create Dumbotas, different in their capabilities and options, which, as their authors hope, in the near future will create the first “Dombot community” that will communicate not only with EIshniki, that is, with us, but also actively communicate with each other .
Therefore, I am deeply convinced that soon we will have to revise the aging “test of Alan Turing” and take, for example, the criterion of Eduard Khachukayev, namely: “A computer program like Dumbot, pretending to“ intellectuality ”, as an artificial mental activity, recognized as such if it, being installed on two different computers (servers), can carry out a dialogue in natural language between them on several topics with an unlimited duration of time, which experts recognize “intelligence ktualnym ".
The level of the well-known "perceptron" technologies (and those of various "neuro-quantum-semantho-statistical-mathematical-network crap") is just a technology of physiology. That is, the signal came to the input, which means that it performed the corresponding action. But these are just “brainless automata” and now they are called “weak AI”.
However, the natural human intellect is based mainly on the second technology - on the Mind. And this is already a "strong AI." That is what their fundamental difference is in Consciousness.
Technology of Mind, like any other technology, contains a Methodology, as a set of Methods, and Toolkit, as a set of Tools.
So, the main method of Reason is THINKING, and the main tool is LANGUAGE.
From this it follows that if we want to repeat the UI of a person in its computer version, like AI, then all we need is nothing - to implement “computer thinking” along with the corresponding software language.
')
If with the latter, like, no difficulties arise, then, here, with the “artificial-thinking iron”, serious problems still arise. It turns out that for the full implementation of the "machine mind", it is necessary for the computer to be able to understand the meaning of the context at the human level, and still be able to operate with a multitude of knowledge.
And, if the software developers were able to saddle the data well, then with knowledge the situation is not important at all. Today there is not even a standard for the format of their storage and processing. That is, the Database is, and the Knowledge Base is not. And what you call the BZ, do not believe in your ears or your eyes. Above the relational database, we practically did not rise.
Worsening the situation and the fact that linguists with cybernetics over the past half century have not been able to come to a common opinion and approach in teaching computers to understand natural languages and extract meaning from the texts written in these languages. And without this, there can be no progress in creating a "strong AI." So, pundits from many academic circles who are involved in this issue have retreated from this very heavy task that has turned out for them.
From here and results, more precisely, their full absence. Therefore, the vast majority of researchers went into the "weak AI", leaving the problem of creating an "Artificial Mind" to their descendants.
True, there are quite a few who stormed this unaffected fortress with great stubbornness. And there are already obvious successes. The world's first Technology Heuristic-Content Processing of Texts (TESL) has been created, called “Ontology”. It is a pioneering program in its class and in its own way, capable of understanding the meaning of the context and even puzzling with questions. That is, glimpses of the first artificial thinking have already appeared. So not everything is so bad on the fabled fronts. At the very least, we can already report on “MEDIUM AI”.
The dialogue system created on its basis has the name "DUMBOT" (from THINKING ROBOT). Today, different developers started to create Dumbotas, different in their capabilities and options, which, as their authors hope, in the near future will create the first “Dombot community” that will communicate not only with EIshniki, that is, with us, but also actively communicate with each other .
Therefore, I am deeply convinced that soon we will have to revise the aging “test of Alan Turing” and take, for example, the criterion of Eduard Khachukayev, namely: “A computer program like Dumbot, pretending to“ intellectuality ”, as an artificial mental activity, recognized as such if it, being installed on two different computers (servers), can carry out a dialogue in natural language between them on several topics with an unlimited duration of time, which experts recognize “intelligence ktualnym ".
Source: https://habr.com/ru/post/149055/
All Articles