Artificial intelligence as a result of human thoughtlessness

The longer I think about what artificial intelligence is, the more I become convinced that this concept is the result of either thoughtlessness on the part of some participants in the scientific process, or a meaningful penetration of others into a financial feeding trough.

Watch my hands: I'm not a sharpie, I'm just reasoning.

What is artificial intelligence? Is it something artificial, yet likened to man, is it not?

If you answer: “So,” then you sign that a person is a set of certain natural algorithms available for understanding and reproduction. Repeat the process of thinking in a mechanical doll, that is, the natural algorithms implemented in the human brain - the doll is likened to a person.

What is the difference between artificial intelligence and the non-intellectual mechanism: an android from some fantastic series from an excavator digging a ditch everyday?

Nothing fundamental, because both operate on the basis of the algorithms embedded in them. Mechanisms always act in accordance with the algorithms laid down in them; they are so arranged.

I know, I know, there is a fundamental difference. Artificial intelligence is required to respond depending on the current state of affairs: when he sees mom, say “mom,” and when he sees dad, say “dad,” it’s important for the non-intelligent mechanism to see who he sees, which program is running.

This, as everyone knows, is called feedback.

Here are a couple of definitions from Wikipedia:
Feedback in technology is a process that leads to the fact that the result of the functioning of a system affects the parameters on which the functioning of this system depends.
Feedback in cybernetics is the presence of circuit cycles in the unchanged part of the machine and conditional instructions in its variable part.

I do not really like these definitions. I would say this: feedback is a change in the system depending on all its previous changes. Hearing outside the door: “Your mother came, milk brought,” a rational creature will open the door, but if in the previous times a gray wolf appeared in place of mom, it will not open. The object of perception is one: someone who calls himself a mother - but the reaction is different depending on previous experience.

It is the presence of feedback that artificial intelligence differs from “non-intellectual” mechanisms.

Is it difficult to construct such a feedback mechanism? There is nothing easier.

Take a calculator. Is it artificial intelligence, that is, according to our definition, a feedback-based mechanism? In no case! There is no feedback in it: you press buttons 2 * 3 and you get 6 as the result, regardless of previous clicks.

We slightly modify the program. Let's make the actions memorized and in case of repetition the message “You already asked” is displayed. I click on the buttons 2 * 3 for the first time, I get 6. I click on the same buttons again, I get the message "You already asked."

This is artificial intelligence - of course, very simplified, but real. A sensible reaction to the changing environment is evident - in full accordance with the algorithm incorporated into the mechanism, otherwise nothing.

What follows from this? The fact that the mechanisms with artificial intelligence are designed by millions! Any database in which new data is alternately added and issued, in accordance with the changed content, answers to queries - this is artificial intelligence, implemented and functioning properly.

And you didn’t want that when you talked about artificial intelligence ?! What? Give you a human, to briefly exchange opinions about female legs and the latest version of the operating system? Do you crave conversations with an artificial being on free topics that is?

Okay, imagine an automaton that responds to your phrases in such a rigidly given sequence: “Solar”, “Not less than you”, “Because of gravity”.

The first person comes up, says: "Hello." The automatic machine answers him: “Solar”. The interlocutor is amazed: "What do you mean?" The automaton answers: "Not less than you." Etc. It is unlikely that the interlocutor will come to the conclusion that he is dealing with a rational being.

The second person comes up, the situation repeats - the phrases vary, but the conclusion about the unreasonability of the automaton remains the same. This is repeated a hundred million times.

And then a hundred million first passer-by approached the machine and asked: “What is the weather like now?” The automatic machine answers: “Solar”. Wai-wai, and the weather is really sunny! Pleasantly surprised counterpart continues to pry: "Are you a rational creature?" The automaton answers: "Not less than you." An almost confident interlocutor throws a third test phrase: "Why do objects fall down?" The machine responds: "Because of gravity." The point is in the hat, one hundred million first interlocutor is convinced of the rationality of the machine.

Alas, despite the successful completion of the Turing test, such a talking automaton is no different from a doll that can only pronounce “mom, dad,” except perhaps with a vocabulary.

But let's imagine that the program of the talking machine is complicated: it implements the principle of feedback and the vocabulary is expanded by several orders of magnitude, as a result, it became possible to discuss the female legs and the latest version of the operating system. The machine is trained to respond to keywords and give an acceptable answer for this interlocutor.

Difficulties in the implementation of such an automaton are possible due to the weak formalization of natural languages ​​and the lack of system resources, but not more. Such systems are darkness, and so what? Thanks to the feedback, the automaton “intelligently” responds to the repeated question with the phrase: “Sorry, you already asked!”, While continuing to act in accordance with the algorithms laid down in it, as without them.

We ask ourselves and we ourselves will answer:

• Will a talking automaton circuit be more complicated than a modified calculator circuit? Yes, of course.
• Will a talking machine be more intelligent than a modified calculator? No, why on earth ?! The extravagant elephant is not a mammal to a greater extent than a field mouse.

The funny thing is that a talking automaton is not fundamentally different even from a person who, according to our agreement, is a combination of natural algorithms. Or do you disagree? A person is a set of algorithms, a modified calculator is also a set of algorithms, therefore, a person and a modified calculator are the closest relatives by reason. They themselves wanted artificial intelligence, get the result from your thoughtless desire: recognition that a person is a mechanism.

Suppose this does not frighten you, you persist: yes, man is a natural mechanism, but give us such an artificial mechanism that it resembles the human one in complexity and reactions.

Let me ask: but in fact, why did you give up such a mechanism, since there is nothing in it except for complexity and “human” reactions? For the sake of the Nobel Ponte? And do not want to construct a machine jumping on rough terrain on one leg and at the same time juggling ten tennis balls? I believe this is no less exciting than creating a humanoid robot. And if such an automatic machine starts baking pancakes with jam during juggling, a non-standard technical problem will turn out. But who needs such a unique?

Okay, suppose you talked me into it. Let artificial intelligence - a mechanism that mimics human behavior, while being comparable in complexity to human.

In this case, artificial intelligence is an artificial person, nothing more. Not any artificial intelligence, namely likened to man.

What follows from this? But here, let's go through the points.

For some reason, it seems to me that to imitate human behavior is possible only on the basis of the same initial data that a person operates on. This means that the receptors of an artificial person must perceive the same external world as we do, otherwise there will be an inevitable divergence in perceptions and, as a result, in behavior (an argument for proponents of the computer mind, who believe that anything can be recreated in a computer environment. If the digital nature of human thinking has not been proved, sorry humbly, you have to use analog).

In this regard, the question arises: how much is it possible to ensure the identity of perception that an artificial person possesses to natural human perception?

It is generally accepted that a person has five senses. Vision and hearing are imitated in technology for a long time and everywhere, with the rest - trouble. Devices that allow you to record and reproduce tactile, taste and olfactory sensations have not been invented - at least they are not sold in stores. This means that the sensations of an artificial person will turn out to be insufficient and defective in comparison with natural human perceptions.

Not scary, - they will object to me. - You can use the biological elemental base, constructing an artificial person from the same material as natural. In any case, the identity of perceptions will be ensured.

But how? Stitching from cadaveric parts - ay, Dr. Frankenstein! - or according to the method of Professor Preobrazhensky? .. Which, however, disavowed his own achievements with recognition:
“You can instill the pituitary gland of Spinoza or some other such goblin and make an extremely tall dog. But what the devil? - asks. Please explain to me why you need to artificially fabricate Spinoz, when any woman can give birth to him at any time. "

The difficulty is not even in one way or another to build a biological creature, but in recognizing him as an artificial person. Is a clone grown from a cell an artificial or natural person? They raised, one might say, artificially, but did not put man-made algorithms inside - is it natural, then?

But if vice versa: naturally grown, but programmed artificially?

In this case, we are completely robots, since all of life circumstances, including artificially initiated ones, for example, the media, programmed us all. Modern media and designed to program! From this point of view, it is not necessary to assemble an artificial person from details — mechanical, biological, or biological — of a sufficiently short exposure by a zombie, after which a natural person turns into an artificial monster.

What do we get? In the case of a mechanical elemental base, it is problematic to achieve human perceptions; in the case of a biological base, it is problematic to recognize a creature as artificial and it is generally not clear what to consider as a criterion of artificiality: an elemental base or successful artificial reprogramming of a subject.

These are flowers in comparison with the problem that occurs when you try to fill the processor of an artificial person with information that fits freely in the human brain!

At what age is a citizen considered capable in civilized countries? From 18 years old? But doesn’t it seem to you that during the indicated period you will have to teach an artificial person to life, exactly how do parents teach a small child? You don’t hope that your handiwork will be connected to the Internet for the night, and the next day you will learn the mysteries of the Universe? If you hope, try to do the same with the month-old son: the chances are that by morning the child will turn into an all-knowing professor, exactly the same. But this is the omnipotent Nature, which we are only trying to imitate - human skills look less preferable!

Hello, how do you feel about constructing an artificial person (a machine jumping on one leg, juggling ten tennis balls and baking jam pies) over twenty years of hard work and filling its processor box with information necessary for forty years achievements by the mechanism of human likeness? Sorry, I didn’t hear the answer ...

Yes, you rightly suspect: I think the Turing test is a logically illiterate design (a machine that we thought was reasonable is considered to be a reasonable machine - as I myself did not guess!), I think it’s expensive brain research projects to create artificial intelligence by ordinary cutting the budget (what we did not see cuts, or what ?!), etc.

This is the time to end.

The summary will be as follows:

• Creating artificial intelligence is not difficult: implementation is achieved through feedback. The number of implemented artificial intelligence systems is not countable.
• Imitation of man in artificial intelligence - that which gives food to countless speculations about the uprisings of machines and other science fiction crap - has no reason. There will be no uprisings of either human-like computers, or furious excavators if they are not programmed accordingly.

But this is in one case: if one recognizes a person as a set of some natural algorithms that are accessible to understanding, this is where the conversation began. You know, I have a different opinion.

The inherent ability of each person to comprehend himself testifies to the duality of our essence: a person not only acts according to the laws of biological mechanics, but also what is commonly referred to as the soul in everyday life - an observation point located outside the physical body. Being a point of observation, the soul cannot know its own device, approximately for the reason that the eye cannot see itself, except in a mirror. For the soul, there is no mirror, so it cannot be copied in a mechanical device. How to copy something that you have no idea?

The mistake lay at the very beginning of our reasoning: that a person is a set of natural algorithms that can be studied and reproduced artificially. And here it is: man is something unknowable, so any copy of it will always be partial and never complete! To reproduce bodily reactions is not a problem (at least at the level of selection of human-like reactions expected from the robot), but the human soul ...

In connection with the transition to the philosophical path, I round off: here they don't like this.

Dixi et animam levavi, as the Latins would say.