Can a car think?
In a series of recent discussions on the topic of AI ( 1 and 2 ), a deeply principled discussion arose: are AI methods able to do something that cannot be made by deterministic algorithms and “where is the intelligence”?
Imitation of physiology
The fact is that the term “Artificial Intelligence” (by the way, gradually replaced by the concepts “intellectual systems”, “decision-making methods”, “data mining”) was initially considered as embracing for a large class of models and algorithms that were supposed to work as well as the human brain (according to the ideas of that time).
These include, for example, the notorious neural networks of all kinds and genetic algorithms.
Summary, statistics and analysis
On the other hand, the many methods of the so-called AI are nothing more than the development of sections of mathematics: statistics, operations research, topology, and metric spaces. These include most of the data mining and knowledge data discovery methods, cluster analysis, the method of group accounting of arguments and other things.
')
These are the methods of so-called inductive inference, when general patterns are derived from the available data.
Rules, logic, conclusion
The third special group can combine methods that try to build general patterns and draw conclusions about specific facts. These are methods of deductive inference, and they are represented: the old as the world syllogistic of Aristotle, the calculus of propositions and predicates, various formal systems and logic. Theories of formal and natural languages, various generative grammars, were immediately attached to the edge.
We see that everything that is usually referred to the term "AI" is trying to imitate or logically solve the imitation problem of human intelligence.
The question arises, what does such a person do that which modern computers, built according to the Babbage principles, do not do yet?
One of the definitions of tasks that AI deals with is: “a task for which there is no algorithmic solution or it is not applicable for reasons of computational complexity ”.
Thus, for example, the task of playing checkers was once the task of AI, and after building a complete model and typing a complete database of unimproved moves, it turned into a task of searching through the information base (see 1 and 2 ).
The tasks of "AI" change over time
Perhaps our children will live in the information world, when many tasks will be solved and new ones will arise - from communication in natural languages ​​to automatic control of all types of technology and mechanisms.
However, when each of us heard the words “artificial intelligence,” we really wanted something different.
We wanted to get a machine that knows how to think , which owns the basic skills of learning, generalization; it is capable, like living organisms, to replace some organs with others and to improve. Everyone read early science fiction, right?
Was there a boy?
So where is the intellect lost? When and why did what we wanted to see become dull matmodels and rather inelegant algorithms?
A couple of lines offtopic. If you defend a thesis with the word “intellectual”, then the members of the council will usually ask you to indicate the place in the system that is intellectual, and to prove WHY it is. This question refers to absolutely "unbelievers."
The fact is that the people who invented everything that the modern “AI” stands on were driven by innovative and revolutionary ideas for that time (in fact, our time differs only in that we have already played enough of it, including using modern computing power)
Example 1 (from the region of the unknowable) .
Neural networks of forward signal propagation with the algorithm of back propagation of error (the so-called back-propagation). This is definitely a breakthrough.
A properly configured network (with cleverly selected inputs and outputs) can learn any input sequence and successfully recognize examples that it has not been taught.
A typical experiment is formulated as follows: 1000 examples, in half of which we learn the algorithm, and check the other. And the choice of the first and second half is done by chance.
It works, I personally taught different NAs different tasks at least 10 times and received normal results, with 60-90% correct answers.
What is the problem of neural networks? Why are they not genuine intelligence?
1. Input data almost always needs to be very carefully prepared and pre-processed. Often tons of code and filters are made to make the data edible for networks. Otherwise, the network will learn over the years and learn nothing.
2. The result of learning NA can not be interpreted and explained. And the expert really wants this.
3. Networks often just memorize examples, rather than learning patterns. There are no exact ways to build a network smart enough to present a pattern and not sufficiently capacious to stupidly remember the entire sample.
What is the intelligence of neural networks?
In that we did not teach the system to solve a problem, we taught it to learn how to solve problems. The algorithm for determining the sex of a person is not incorporated into the system by a person; it is found almost empirically and is sewn into the weights of synapses. This is an element of intelligence.
Example 2 (from the field of deductive inference).
The idea is simple. We will train the car to think like a person (well, at least to make primitive conclusions) and give basic facts. Next - let her.
Expert systems, computer logic systems, ontologies (with some stretch) work according to this principle. It works? Of course. Thousands of disease diagnostic systems and descriptions of knowledge areas are implemented and continue to work.
What is the problem? Why are formal systems not genuine intelligence?
The problem is that the system, having absorbed the colossal amounts of blood and sweat of its creators, at the very least begins to repeat and develop the solutions of the expert (or the community) who taught it.
Is it helpful? Undoubtedly. The expert is mortal, tasks are multiplying.
What is the intelligence of knowledge-based systems?
The fact that the car makes new conclusions that no one has taught it. This element of her work is extremely poor (so far) and limited to those models and algorithms that were laid. But it is - an element of intelligence.
So what is the problem of modern AI?
Simply, we are still very small. Our naive and superficial ideas about how a person thinks and how the brain works give the fruits they deserve.
We are certainly incredibly far from creating machines that would be able to think in our human sense, but our steps in this direction are correct and useful.
And even if we are not going there, who knows, maybe like the Strugatskys, we will inadvertently do something better as a result of directed efforts than we were going to?
Links
0. Can cars think? Alan Turing
1. Can cars think? Leonid Ashkinazi.
Imitation of physiology
The fact is that the term “Artificial Intelligence” (by the way, gradually replaced by the concepts “intellectual systems”, “decision-making methods”, “data mining”) was initially considered as embracing for a large class of models and algorithms that were supposed to work as well as the human brain (according to the ideas of that time).
These include, for example, the notorious neural networks of all kinds and genetic algorithms.
Summary, statistics and analysis
On the other hand, the many methods of the so-called AI are nothing more than the development of sections of mathematics: statistics, operations research, topology, and metric spaces. These include most of the data mining and knowledge data discovery methods, cluster analysis, the method of group accounting of arguments and other things.
')
These are the methods of so-called inductive inference, when general patterns are derived from the available data.
Rules, logic, conclusion
The third special group can combine methods that try to build general patterns and draw conclusions about specific facts. These are methods of deductive inference, and they are represented: the old as the world syllogistic of Aristotle, the calculus of propositions and predicates, various formal systems and logic. Theories of formal and natural languages, various generative grammars, were immediately attached to the edge.
We see that everything that is usually referred to the term "AI" is trying to imitate or logically solve the imitation problem of human intelligence.
The question arises, what does such a person do that which modern computers, built according to the Babbage principles, do not do yet?
One of the definitions of tasks that AI deals with is: “a task for which there is no algorithmic solution or it is not applicable for reasons of computational complexity ”.
Thus, for example, the task of playing checkers was once the task of AI, and after building a complete model and typing a complete database of unimproved moves, it turned into a task of searching through the information base (see 1 and 2 ).
The tasks of "AI" change over time
Perhaps our children will live in the information world, when many tasks will be solved and new ones will arise - from communication in natural languages ​​to automatic control of all types of technology and mechanisms.
However, when each of us heard the words “artificial intelligence,” we really wanted something different.
We wanted to get a machine that knows how to think , which owns the basic skills of learning, generalization; it is capable, like living organisms, to replace some organs with others and to improve. Everyone read early science fiction, right?
Was there a boy?
So where is the intellect lost? When and why did what we wanted to see become dull matmodels and rather inelegant algorithms?
A couple of lines offtopic. If you defend a thesis with the word “intellectual”, then the members of the council will usually ask you to indicate the place in the system that is intellectual, and to prove WHY it is. This question refers to absolutely "unbelievers."
The fact is that the people who invented everything that the modern “AI” stands on were driven by innovative and revolutionary ideas for that time (in fact, our time differs only in that we have already played enough of it, including using modern computing power)
Example 1 (from the region of the unknowable) .
Neural networks of forward signal propagation with the algorithm of back propagation of error (the so-called back-propagation). This is definitely a breakthrough.
A properly configured network (with cleverly selected inputs and outputs) can learn any input sequence and successfully recognize examples that it has not been taught.
A typical experiment is formulated as follows: 1000 examples, in half of which we learn the algorithm, and check the other. And the choice of the first and second half is done by chance.
It works, I personally taught different NAs different tasks at least 10 times and received normal results, with 60-90% correct answers.
What is the problem of neural networks? Why are they not genuine intelligence?
1. Input data almost always needs to be very carefully prepared and pre-processed. Often tons of code and filters are made to make the data edible for networks. Otherwise, the network will learn over the years and learn nothing.
2. The result of learning NA can not be interpreted and explained. And the expert really wants this.
3. Networks often just memorize examples, rather than learning patterns. There are no exact ways to build a network smart enough to present a pattern and not sufficiently capacious to stupidly remember the entire sample.
What is the intelligence of neural networks?
In that we did not teach the system to solve a problem, we taught it to learn how to solve problems. The algorithm for determining the sex of a person is not incorporated into the system by a person; it is found almost empirically and is sewn into the weights of synapses. This is an element of intelligence.
Example 2 (from the field of deductive inference).
The idea is simple. We will train the car to think like a person (well, at least to make primitive conclusions) and give basic facts. Next - let her.
Expert systems, computer logic systems, ontologies (with some stretch) work according to this principle. It works? Of course. Thousands of disease diagnostic systems and descriptions of knowledge areas are implemented and continue to work.
What is the problem? Why are formal systems not genuine intelligence?
The problem is that the system, having absorbed the colossal amounts of blood and sweat of its creators, at the very least begins to repeat and develop the solutions of the expert (or the community) who taught it.
Is it helpful? Undoubtedly. The expert is mortal, tasks are multiplying.
What is the intelligence of knowledge-based systems?
The fact that the car makes new conclusions that no one has taught it. This element of her work is extremely poor (so far) and limited to those models and algorithms that were laid. But it is - an element of intelligence.
So what is the problem of modern AI?
Simply, we are still very small. Our naive and superficial ideas about how a person thinks and how the brain works give the fruits they deserve.
We are certainly incredibly far from creating machines that would be able to think in our human sense, but our steps in this direction are correct and useful.
And even if we are not going there, who knows, maybe like the Strugatskys, we will inadvertently do something better as a result of directed efforts than we were going to?
Links
0. Can cars think? Alan Turing
1. Can cars think? Leonid Ashkinazi.
Source: https://habr.com/ru/post/51521/
All Articles