"The problem with the artificial intelligence system": The strike UAV "Okhotnik" as an interceptor has little chance in front of a manned fighter

Russia continues to improve the first domestic heavy shock drone "Hunter". The tests carried out by the developer allow not only to clarify (confirm) the technical characteristics of the UAV, but also to obtain comprehensive information about the options for using this drone as part of the Russian Aerospace Forces.

One of the directions for the future use of the Okhotnik UAV, as noted, is associated with the breakthrough of the enemy's air defense system. Moreover, an attack drone can use several types of weapons: air-to-surface missiles to destroy air defense systems on the ground, air-to-air missiles to destroy air targets - aircraft that can be sent to intercept.

At the same time, the statements that the attack drone itself can be used as an interceptor - in the presence of air-to-air missiles - cause some skepticism among experts. It is associated with the drone control functionality. If at the moment (during testing) it is controlled from the ground by an operator and if this option remains the basis for some time after the Hunter is put into service, then the probability of success in air combat against a manned fighter will be low. After all, the pilot in the cockpit of the fighter and the operator in front of the monitor, by definition, have different volumes of data on the operational situation in the air in case an optimal decision is made.

It will be another matter if the "Hunter" in the end turns out to be completely robotic. But in this case, too, there are still hypothetical problems. They are connected with the robotization system itself. If this is partial robotization, for example, in choosing the optimal flight route, then there is also little chance as an interceptor in front of a manned fighter. If this is a complete robotization with artificial intelligence, then the question is different. The fact is that so far none of the world's leading companies in the creation of AI systems guarantees that in the near future artificial intelligence in military equipment will unambiguously surpass a person (for example, a pilot of a combat aircraft) and will be able, after rationally assessing the situation, to strike at the right time for the desired goal. The reason lies in the still technical limitations of AI systems. What's great on the chessboard can be a losing proposition in real air combat. At least for the time being and for the near future.

Therefore, at this stage, it should be noted that for the functions of an interceptor, the strike UAV "Hunter" requires advanced hardware and software for it, so that the robotization of the drone itself does not become a formality. But in order to put this advanced hardware with suitable software on shock drones, we need our own production. It is definitely impossible to rely on manufacturers in Europe or Southeast Asia in this regard, no matter how trite it may sound.

Comments

[A009]

As far as I remember, recent exercises in the United States have shown the total superiority of AI over pilots.
This is to the statement that: "... The fact is that so far none of the world's leading companies in the creation of AI systems guarantees that in the near future artificial intelligence in military equipment will definitely surpass a person (for example, a pilot of a combat aircraft) .. . "
Are you, the author, sure about your statements?

[NIKN]

experts cause a certain skepticism. It is associated with the drone control functionality.

Yes, there is a problem, but it can be solved. The fact is that in terms of the amount of information, the ground network is more informed than the pilot in the cockpit. Here the question is by what means to control the means of destruction. For today, I suppose, you can direct a rocket using ground means or air (it does not matter) here and AI rules. Of course, the pilot drives the BVB, but before him .... you still need to get to the BVB alive.

[umah]

Nobody forces the "Hunter" to independently search for air targets and conduct close maneuvering combat. The enemy is detected by ground (or other) means, the drone is brought to the launch range of air-to-air missiles, and a command is given to launch.

And the comments of the "experts" as always: analysis of a spherical aircraft in a vacuum - there are no AWACS aircraft, there are no ground radars, there is nowhere to get target designation ...

[Peak]

Quote: umah

Nobody forces the "Hunter" to independently search for air targets and conduct close maneuvering combat.

And who, in general, makes us discuss some statements of anonymous experts who have stated what they themselves have not seen, and probably have no idea what the "Hunter" will be like in the end?

[venik]

Quote: PiK

And who, in general, makes us discuss some statements of anonymous experts who have stated what they themselves have not seen, and probably have no idea what the "Hunter" will be like in the end?

=======
And "talk" ??!

[Paranoid50]

Quote: PiK

And who, anyway, makes us discuss some statements of nameless experts

However, the "dear editorial office" went haywire, shoving anonymous letters to the tape out of bounds.

[Puler]

Taking into account that the exact information on the Hunter is likely chipboard .... and never discuss this kind of work of a combat unit on the network without waking up ... a throw on the fan.

[VO3A]

You read all this nonsense in the article and in the comments and you are more and more convinced that first they make an airplane or UAV, and then figure out how to use it .. A country of fools with a field of miracles ... This applies to both the Su-57 and the Hunter. , and an even more popular bunch ...
Let's go back to the sinful earth. The main task of the UAV is reconnaissance and obtaining information in a covert mode and transmitting it in real time, even without the risk of destroying the pilots, even in the face of opposition and the presence of subtle hidden enemy reconnaissance means ...
We sculpt exactly the opposite, forgetting the main thing: where to get information about the enemy in real time and ensure its destruction in real time in a rapidly changing environment?
The most real application of the "Okhotnik" in conjunction with a conventional Su-30 aircraft with two pilots, one of which performs the task as an operator for monitoring and control of a robotic UAV, and analyzing the incoming information from it in real time ... At the same time, the UAV provides a hidden detecting air defense systems and enemy aircraft and transmitting information, including on board the Su-30, which is located with weapons outside the detection zone and fires and destroys the enemy on the ground and in the air according to target designation signals from the UAV ... identification of a foreign aircraft and duplication of the friend-or-foe system, which is the weakest link in the chain of decision-making at a long distance to a target in a long-range launch in real time, and making a decision on "start-up"... Here is the simplest typical and highly demanded use case ...

[bar]

Quote: A009

As far as I remember, recent exercises in the United States have shown the total superiority of AI over pilots.

On what tasks did they show this superiority? When delivering assault strikes against ground targets? Or when launching air-to-air missiles from a long distance? I will never believe that AI can win in real air combat. Exercises in the USA are still exercises, there the pilots themselves do not condescend to "dog dumps", and their planes are not sharpened for this. They have a paradigm to sneak up, launch a rocket and quickly dump. Here, a drone with an AI can, of course, surpass the pilot in stealth and stall speed.

[ugol2]

Quote: bar

the pilots themselves do not condescend to "dog dumps"

Somehow you are pejorative about the pilots.
Excuse me, but for the infantry "dog dump" ie. Is hand-to-hand combat the main type of combat? Or, due to technical progress, they moved away from this 500 years ago, and now they prefer to destroy each other remotely from shelters, using cannons, machine guns, etc.?
Aviation follows the same path.
Mastering the techniques of hand-to-hand combat or a dog dump in relation to aviation, of course, is a big plus.
But the modern army differs from the warriors of the Tumba-Yumba tribe in that it wins primarily due to its advanced technical equipment. And in hand-to-hand combat (dog dump), the big question is who will keep the battlefield. Most likely for the evil natives from the aforementioned tribe.

[bar]

Quote: ugol2

Excuse me, but for the infantry "dog dump" ie. Is hand-to-hand combat the main type of combat? Or, due to technical progress, they moved away from this 500 years ago

Excuse me, but now how? How does it come to hand-to-hand combat, do the opponents collect their things and go back to the original ones?

Aviation follows the same path.

Aviation follows different paths. Someone relies on "invisibility" to the detriment of flight performance, and someone, on the contrary, on super-maneuverability. Only a real war can show who is right and who will win. In the meantime, you can dream about AI.
But here such a thing - for example, automakers, who stuffed modern cars with all sorts of intellects, are now in the anus with production. The Asian tigers have failed in the supply of processors and electronic components due to covid. How AI will survive in the event of a war is a big question. But this is so, thinking out loud ...

[ugol2]

You don't seem to read what I am writing:

Mastering the techniques of hand-to-hand combat or a dog dump in relation to aviation, of course, is a big plus.

BUT:

Is hand-to-hand combat the main type of combat?

The answer is no.
Remotely. The army has been trying to fight remotely for 500 years. Hand-to-hand combat is an exception when it doesn't work out otherwise.
Aviation cannot have any development path of its own. She also has to fight remotely. It is perfectly natural to try to destroy the enemy at the maximum distance, and at the same time not to incur losses yourself.
I would not contrast the super-maneuverability of stealth. (It doesn't occur to you to compare a dagger with a grenade launcher) First, stealth and DVB. Well, if everything is bad, then we take out the "dagger" and go into hand-to-hand combat with super-maneuverability ... Only for this we must still remain alive, therefore stealth is in the first place.

[A009]

Right, I don't even want to comment. There is everything available. Even the analysis is quite detailed, with good analyzes and alternatives for conclusions. And the AI ​​won more than confidently. May I ask you what is more difficult air combat or striking the ground?

[Yuriy Filatov]

And what is the difference between a pilot and an AI in your opinion?
The pilot operates with exactly the same data as the AI, only it takes much longer to make decisions and is much more sensitive to overload.

[Dude]

As far as I remember, recent exercises in the United States have shown the total superiority of AI over pilots.

Colleagues, I don't understand in any way what is being discussed now? What total or specific superiority AI can, today, be talked about?
AI is a matter of the future, perhaps the near future. But, at the moment, artificial intelligence systems, simply, There is no.
Anything that is called this loud term today is not AI, by definition.

[lis-ik]

Quote: Dude

Everything that is today called this loud term, AI is not, by definition

And in the near future, too, is unlikely.

[Dude]

Quote: lis-ik

Quote: Dude

Everything that is today called this loud term, AI is not, by definition

And in the near future, too, is unlikely.

We'll see. So far, even with the Turing test, machines are not doing very well, although Turing himself believed that this would happen by 2000.

[Pandiurin]

Quote: lis-ik

Quote: Dude

Everything that is today called this loud term, AI is not, by definition

And in the near future, too, is unlikely.

It's all about terminology and AI means different things.

1. As in science fiction novels or films Artificial intelligence that has come to self-awareness. This is the most complete implementation of AI. And in the short term it is not achievable and generally not interesting for applied developers.

2. Certain algorithms with elements of artificial intelligence for practical tasks. In such cases, they always speak in conjunction, which gives an understanding for which specific task this AI is sharpened. For example, a car with AI autopilot.

The second option is already today.
Many tasks are performed using such systems.
So far, the implementations have quite a few drawbacks, but they are already there today.
Such systems are getting better every day. There is hardware (processors) under them.
And the main thing is that now there is huge funding for this area and large teams are working.

AI (even insufficiently perfect ones) will be installed on fighters, if only due to the fact that, without a pilot, such a machine can maneuver more than 10G.
AI is a piece of hardware with loaded software, it does not require training for several years as for a human pilot, this is the second argument.

Now a variety of unmanned drone systems are being developed, do the warriors really want to control them with the help of an operator on a joystick, it is obvious that it is supposed to get an advanced system, which is possible only if there is a bius with AI.

[Simple]

Quote: Pandiurin

AI is a piece of hardware loaded with software, it does not require training for several years as a human ...

AI is software. And it just requires training.
To "learn", Soft needs information on the topic of interest (AI). The more examples you load into the AI ​​(for the most part, "fresh" - before the flight) - the better it (AI) will build algorithms.

In other words, learning for AI is no less important than for humans.

[Pandiurin]

Quote: Simple

Quote: Pandiurin

AI is a piece of hardware loaded with software, it does not require training for several years as a human ...

AI is software. And it just requires training.
To "learn", Soft needs information on the topic of interest (AI). The more examples you load into the AI ​​(for the most part, "fresh" - before the flight) - the better it (AI) will build algorithms.

In other words, learning for AI is no less important than for humans.

If there is already a trained AI, for example, 10.000 flights were carried out on simulators, simulated battles, and let's say a certain number of real combat missions.
That experience is presented in the form of a base of values, add. Algorithms, settings / states for a processor simulating a neural network. And all this can be backed up. This is a necessary feature because when the power is turned off or the unit is replaced, the AI ​​should not be reset to kindergarten.

Unlike human brains, all hardware is absolutely standard on one drone or another.
Therefore, the software is simply copied, the base with "experience" is copied and we will get a fully trained AI.

Training in commercial products is required, for example, for a new user, for example, for voice recognition, because Petrov, Ivanov and Sviridova have individual characteristics that the manufacturer does not know and cannot do this in advance because all buyers are, in principle, different.

It also does not require setting up training and developing a knowledge base for a specific "user". Just a lot of experience with a variety of situations is used. Therefore, the "experience" is general or several options that have shown themselves well, in order to exclude the situation that all drones perform some certain predictable sequence of actions.

[Simple]

Quote: Pandiurin

If there is already trained AI, for example, 10.000 flights were carried out on simulators, simulated battles, and let's say a certain number of real combat missions ... (and further down the list)

That is, you do not deny the need to train AI and provide the latter as much as possible (necessary to complete the set)?

And yes - the "experience" for the AI ​​comes during its (AI) training.
The question is, why then did they comment on my comment?

[Yuriy Filatov]

Quote: Simple

In other words, learning for AI is no less important than for humans.

True, but the trick with AI is that having trained one copy, you can serially churn it out, you can change its behavior progamatically for certain tasks before departure. But this won't work with a pilot.

[Simple]

Quote: Yuri Filatov

True, but the trick with AI is that having trained one copy, you can serially churn it out, you can change its behavior progamatically for certain tasks before departure. But this won't work with a pilot.

All right.

But I also did not question it.

We with the members of the forum did not fully understand each other.

[ProkletyiPirat]

Quote: Pandiurin

It's all about terminology and AI means different things.

Everything is much worse, now there is no such thing as "intelligence" under this word in different segments of the info-field mean different things \ processes \ phenomena.

In general, there are many classifications of intelligence in general and AI in particular. In my opinion, the most accurate description of intelligence is used in anthropology and biology, it sounds like "intelligence is the ability to solve non-standard problems with non-standard methods." That is, if there is a standardization of tasks / methods, then this is no longer intelligence, but something else. This other is called "Self-Optimizing Computing Modules (SOCM)" they can be both software and hardware, as well as software and hardware.

[Pilat2009]

Quote: A009

As far as I remember, recent exercises in the United States have shown the total superiority of AI over pilots.
This is to the statement that: "... The fact is that so far none of the world's leading companies in the creation of AI systems guarantees that in the near future artificial intelligence in military equipment will definitely surpass a person (for example, a pilot of a combat aircraft) .. . "
Are you, the author, sure about your statements?

And what did they prove? Yes, the volume of information processing in AI is higher, but in order to capture a target and launch outstanding intelligence, you do not need to. The main purpose of the drone is to patrol on the border of air defense action, monitor the situation and release missiles that are guided by themselves. he doesn't care

[ProkletyiPirat]

Quote: Pilat2009

the amount of information processing in AI is higher

this is not so, just AI \ SOVM knows how to structure and optimize input / output data, the human brain "can" but does not "know how", For example, if you and I had the ability to control the neurons of the brain, we could turn part of the brain into " calculator "and quickly solve the tasks on the basis of structured data, but there is no such technology, but there are" unique ones "who have such neuro-structures formed naturally and such people can multiply five to ten-digit numbers in their minds.

Quote: Pilat2009

the main purpose of the drone is to patrol on the border of the air defense action, monitor the situation and release missiles that are guided by themselves

you are confusing cause and effect. no one bothers to make a drone for other purposes, the only question is the necessity and rationality of this action ..

[Narak-zempo]

After all, the pilot in the cockpit of the fighter and the operator in front of the monitor, by definition, have different amounts of data on the operational situation in the air in case an optimal decision is made.

What is the problem to give the operator an augmented reality helmet instead of a monitor, so that he feels like he is sitting in a "glass cockpit", and transfer all data about the tactical situation there?
On the contrary, the operator will have an advantage due to the absence of fear, the influence of overloads, etc.
Of course, in the presence of a "thick" and reliably working communication channel.
But with this we are all right, as shown by the tests of "Uran-9".

[Victor_B]

Quote: Narak-zempo

What is the problem to give the operator an augmented reality helmet instead of a monitor, so that he feels like he is sitting in a "glass cockpit", and transfer all data about the tactical situation there?

Delay in SECONDS (!) For the passage of information and back commands.
That is, the concept of "real-time control" is lost.
Undoubtedly, the Hunter is controlled remotely and the operator sees the situation through cameras.
But the "thick" video communication channel cannot be transmitted without loss from the maneuvering platform.

[Narak-zempo]

Quote: Victor_B

Delay in SECONDS (!) For the passage of information and back commands.

Where does such a delay come from, dear?
We do not control the device on the Moon when the signal goes to one end for more than a second.

[Victor_B]

Quote: Narak-zempo

Where does such a delay come from, dear?

Because for a long time control and communication are not analog, but digital!
Digital data processing takes longer than analog data, and if the signal goes through communication satellites, then the time increases even more.
Just watch live TV broadcasts over the Internet and on local TV.

[military_cat]

Quote: Victor_B

Just watch live TV broadcasts over the Internet and on local TV.

What happens in modern means of digital signal processing - compression, adaptive change of parameters, noise-immune coding, encryption - cannot be realized using analog methods. Therefore, it is not worth comparing the incomparable. At the same time, if you simply repeat digitally the manipulations performed with the signal in analog systems, they will work just as quickly (only no one needs an unencrypted unencrypted channel of excessive width and power in 2k21, except for TV broadcasts, although it already dies off even there). Where a minimum delay is required in digital (not in TV broadcasts), it is provided.

[ProkletyiPirat]

Quote: Victor_B

Quote: Narak-zempo

Where does such a delay come from, dear?

(1)Because for a long time control and communication are not analog, but digital!
(2)Digital data processing takes longer than analog data, and if the signal goes through communication satellites, then the time increases even more.
(3)Just watch live TV broadcasts over the Internet and on local TV.

1) And the analog method of data transmission apparently has no delays
2 + 3) well, yes, yes, to confuse data transmission by different methods, channels, and even to confuse processing with transmission is generally "the humanities of the brain".

[Glory1974]

Delay in SECONDS (!) For the passage of information and back commands.
That is, the concept of "real-time control" is lost.

In close combat, this is probably essential. And at a distance of tens of kilometers, it no longer plays a role. Already now the plane is largely controlled by a computer and the pilot only chooses actions from the proposed ones.

[bar]

The conversation seems to be for artificial intelligence, and not about remotely controlled devices? Or does the AI ​​not work without props in the form of an operator? Duc kagbe's article is about this.

[Victor_B]

Quote: bar

The conversation seems to be for artificial intelligence, and not about remotely controlled devices? Or does the AI ​​not work without props in the form of an operator?

Surely!

[Roman070280]

Duc kagbe's article is about this.

The kagbe article is about everything ..

If at the moment (during testing) it is operated from the ground by the operator and if this option remains for some time after the Hunter was put into service, the probability of success in aerial combat against a manned fighter will be low. After all, the pilot in the cockpit of the fighter and the operator in front of the monitor, by definition, have different amounts of data on the operational situation in the air in case an optimal decision is made.

Another thing, if the Hunter ends up being fully robotic. But in this case too

[bar]

Quote: Roman070280

It will be another matter if the "Hunter" in the end turns out to be completely robotic.

Even if the "hunter" is fully robotic, he is unlikely to be better than the pilot in real air combat. Another thing is striking ground targets.

[Roman070280]

I just answered the question

The conversation seems to be for artificial intelligence, and not about remotely controlled devices?

[Pandiurin]

Quote: bar

Quote: Roman070280

It will be another matter if the "Hunter" in the end turns out to be completely robotic.

Even if the "hunter" is fully robotic, he is unlikely to be better than the pilot in real air combat. Another thing is striking ground targets.

It's like with chess (although there is no AI) computers play worse than grandmasters (there are also at the level) but much better than masters of sports.
Here, it is also possible that there will be unique aces pilots, but there are only a few such pilots, and they are not replenished. And the AI ​​is consistently much higher than the average.

Again, according to the formulation of the task during the attack
1 by 5, AI is by definition better in being a "hero".
Of the two options
- attack from afar and leave,
with a probable score of 0: 1
- cut to the fullest,
with a probable score of 1: 4
no doubt the second option will be chosen.

[Antidote]

As shown by the tests of "Uran-9", with this we are just not all right

[iouris]

There is nothing more complicated than intelligence. AI is not a "system", but a complex system of the highest level of complexity. There will always be problems with AI as long as they try to create it. The intellectual and psychophysiological capabilities of a person are degrading, therefore, sooner or later, the automatic control system will begin to outplay the person.

[Narak-zempo]

Quote: iouris

There is nothing more complicated than intelligence

I think the intelligence of the average conscript is not that difficult.

[Glory1974]

I think the intelligence of the average conscript is not that difficult.

But he is not predictable!

[iouris]

Quote: glory1974

But he is not predictable!

Oh, is it? The most difficult type of intellectual activity today is filling out a ballot and choosing something in a store. And this is very predictable.

[Herman 4223]

It should aim at an air target on command from a Su-57 fighter. He's part of the pack.
He doesn't seem to need artificial intelligence for a completely independent flight.

[loki565]

What prevents it from being used in conjunction with airplanes? Even as a distracting target, fly in front and provoke the enemy to launch missiles, and the planes will already hit the revealed enemy. If we finally finish the fire-and-forget missiles, then the UAV itself can stand up for itself.

[kytx]

the Americans on the simulator fought a live pilot in f15 with a car also in f15
the car won 5: 0
back in the 90s in computer simulators
type F15StrikeEagle or F16FalkonGold for example
the computer was quite good at dogfight
the physics of flight in these games was quite on the level, overloads were attached :)
it all worked on a dead x286

[Glory1974]

the Americans on the simulator fought a live pilot in f15 with a car also in f15
the car won 5: 0

Previously, the computer could not play chess either. Now he does not give a chance to any grandmaster.

[kytx]

in chess there are a finite number of games, the computer plays chess in a completely different way not like a person. air combat is something else.
the truth of real AI has not yet been created

[bar]

Quote: kytx

the Americans on the simulator fought a live pilot in f15 with a car also in f15
the car won 5: 0

Not surprised. The striped ones have long been degrading in terms of natural intelligence. And their new aged president is a vivid confirmation of this. I'm not talking about the president himself, but about those who elected him.

[kytx]

amers and NATO members in general have excellent fighter pilots with courage - everything is right. don't kid yourself

[El Chuvachino]

Nonsense with false theses written by a person far from the topic.

[Runoway]

... But in order to put this advanced "hardware" with suitable software on shock drones, you need your own production

It is necessary, very necessary, yesterday, and we have one aliexpress

[Vladimir_2U]

At the same time, statements that the attack drone itself can be used as an interceptor - in the presence of air-to-air missiles - cause some skepticism among experts.

Interception and air combat were clearly confused, back in the 60s. the interception was carried out by commands from the ground and the pilot only carried out commands, not even voice ones.

[Wedmak]

And the author does not confuse interception and air combat? We have an excellent interceptor MIG-31BM, how it actually works: takeoff, entering the intercept course, fixing / or identifying the target, if you need a volley. Can't a remotely controlled drone handle this? And given his ability to stay in the air for hours, patrolling the area - what is the problem of interception?

[Hexenmeister]

what is the problem of interception?

For the guidance of long and medium-range V-B missiles, at a certain stage of their guidance, it is necessary to transmit radio correction signals to the missiles. The MiG-31 has a radar in its nose, tracking targets, with a good antenna and not sickly power, but what is in the nose of this "Hunter" ??? If he simply launches a rocket, then there must be a fighter nearby, whose radar will transmit these signals, and he must be close, the potential of this line, especially for long-range B-B missiles, is already at its limit. And if "everyone is on the drone", then there must be a good antenna and a powerful transmitter, but so far nothing is visible there. And the tasks of reaching the desired area, ranking targets for an attack, are nothing complicated.

[Wedmak]

I think it will not be difficult to add the radar station. At least as one of the versions of the drone. It is unlikely that it will be fast, but you should not refuse such an opportunity. Patrolling the same SMP is cheaper and easier on drones than driving heavy MIGs.

[Hexenmeister]

A radar is primarily an antenna, and a good antenna is dimensions, and not just anywhere, but in the bow, plus the dimensions of the rest of the radar, and of high energy consumption. And the picture presented in the article does not give me grounds that with such a "look" of the front part, it will have a good radar.

[Castro Ruiz]

Radar on drones will appear only with ROFAR.

[Conjurer]

One of two things: either stories that a support system (co-pilot) has been implemented (is being implemented) for the SU-57, which provides automatic recognition, selection and capture of a target for defeat - a simple idle talk, or (by installing such a system on the Hunter) you will receive an automatic interceptor ... Not for melee, of course, but why do you need melee?
And, in general, artificial intelligence is just a way of processing information. It consists in creating, as a result of the so-called "training" of certain templates, with which the newly received information is then compared. That is, instead of logical analysis - what is in front of you, you simply compare with a template - similar or not similar. The volume of calculations is huge, but programmers do not need to puzzle over the logical analysis algorithm.
This is what I mean, there is no need to exaggerate the importance of AI. These are not brains, this is just a method of mathematical processing.

[Guru]

To put it mildly, a STRANGE article.

At the same time, statements that the attack drone itself can be used as an interceptor - in the presence of air-to-air missiles, experts cause a certain skepticism.

Which Experts , at VO we are all experts of the Divan Troops.
and further in the text:

The fact is that so far none of the leading companies in the world on the creation of AI systems does not guarantee that in the near future artificial intelligence in military equipment will unambiguously surpass a person (for example, a pilot of a combat aircraft) and will be able to rationally assess the situation, strike at the right time on the right target

That is, we take an owl and stretch it over the globe.
That is, the decision to destroy or intercept the target is made by the MAN, but the process itself is performed by the UAV. AI should not in any case independently make decisions on the destruction of a person. Even if he (AI) can do it.

It is definitely impossible to rely on manufacturers in Europe or Southeast Asia in this regard, no matter how trite it may sound.

And with what fright you decided that we have components from Europe or Southeast Asia. We (Russia) have been doing our own for a long time, piece by piece but OWN.

[amr]

Quote: A009

As far as I remember, recent exercises in the United States have shown the total superiority of AI over pilots.

Are you, the author, sure about your statements?

were the exercises simulated on a computer?))

[the same doctor]

Conventional computer architecture is not well suited to control either a fighter jet or a machine tool or automatic transmission. Even the hurricane speed of computations cannot cope. Meanwhile, you can make a computer that will control the plane better at a frequency of 100 hertz ... And at frequencies in kilohertz, there are no problems at all.

[Rubi0]

Shouldn't it become a spaceship? A couple of years ago, they were glad that at least something was being developed in the class of shock reconnaissance drones, they did not have time to accept the new functions of the fighter.

[Castro Ruiz]

Dumayu, in conjunction with the SU-57 with his own 1 - 2 drones or a flock of drones, needs an operator, e.z. two-seater version of the car.

[Knell wardenheart]

The current battles in the air are not "aerobatics" of the times of BB1 - radars, intelligent missiles of various "fire and forget" systems, stealth coverage, various sensors - the robot sees an enemy aircraft at a distance just as the pilot sees it in the cockpit-point on the screen ... To train a neural network to determine the assumptions of what this point is is not a matter of what task, the pilot in this case also operates with not such a huge amount of knowledge and not a heap of time. Further, machine algorithms are already working - the optimal distance for launching missiles is determined by a computer based on data from sensors, the missile will be quite intelligent in any case - the task of modern air combat is thus reduced to good sensors and a good missile / several missiles for different situations - but by no means to the mythical "AI power".

[Guru]

Knell Wardenheart (Knell)
Today, 13: 32
And I'm talking about the same, Decisions are made by a person-Stupid in the YES_NO mode, and the rest is done by the Drone, while the drone itself should not make decisions in any way.

[Knell wardenheart]

Well, I'm talking about the fact that as an interceptor within the borders of the state, the drone is already quite combat-ready within the framework of the realities of modern combat (if the need arises). In this case, the disadvantages of AI can be leveled by directive control, the disadvantages of sensorics - by interaction within a group of drones and between drones and ground-based air monitoring facilities.
In this case, the drone acts as a highly maneuverable and sensor-equipped air-to-air launch platform, and this is our future in reality, since the enemy is superior to us in numbers and demographics, in any case, we will develop robotic systems for defense, and in in any case, we will expand their "rights".

[Shahno]

Quote: Intruder

And the only advantage of AI is the number of these options in memory. But ... But, there is one but. In the case of an AI test in a situation with imperfect information, in other words, not very complete, and with a volume exceeding the number of chess pieces or cards in the deck, but tried on Starcraft, GLA and the fog of war, and dozens of opponents, the AI ​​showed itself a little worse than nothing ... In general, losing quite to itself the average players. To whom no one pays a salary for the game, and even more so does not prepare for it for years. yes So ... Real AI, and not its imitations, is still sooooo far away.

You are talking about arrays of datasets, for training the simplest NN, Duc, they are the most elementary primitive, there are also more advanced, but so far resource-intensive methods are in deep learning and there is a lot of NE hardware support there, the result of NN work depends!

Intruder. It's really cool.

[Alexfly]

Well, that's right, from the point of view of a nationalist .. It was THEY who defeated the German and fought to the death with the NKVD for independence .... Only, a discrepancy, the participation of the OUN and So in the executions and destruction (physical) of the civilian population on ethnic grounds is documented, as well as prisoners of war. It is clear who is putting the spokes in the wheels. the Nazis are not going to shoot in their own leg ..

[Zaurbek]

Judging by the "wing" scheme and a small fairing of a small area .... this is a stealth bomber .... an analogue of F117. And he can most likely launch missiles in the military as an "arsenal", according to the external tsu from machines of the Su30 / 35/57 type ....

[pba50]

Quote: PiK

Quote: umah

Nobody forces the "Hunter" to independently search for air targets and conduct close maneuvering combat.

And who, in general, makes us discuss some statements of anonymous experts who have stated what they themselves have not seen, and probably have no idea what the "Hunter" will be like in the end?

[Ruslan Sledkov]

I do a little programming on the topic of performing engineering calculations. Not a professional. But from experience I can say that what you wrote in the program, it will do it. No more, no less. “Learning” AI can consist of adding blocks to the program that expand the capabilities of the program, rather than teaching it to think. The choice of the optimal solution is also rigidly programmed and is not an evolution of the "thinking" of the program. Maybe now there is something different from what I wrote? May be. I do not know. Enlighten who knows.

[Sancho_SP]

Learning means the development of refinement coefficients for a relatively rough initial program.

[tarragon]

When air-to-air missiles are launched, approximately two-thirds of the missile's path needs to fly after it to direct it to the target. It is not difficult to imagine that in a meeting engagement with simultaneous detection, opponents fly to meet each other's missiles, directing their missiles. What is this distance? Somewhere I came across data on the analysis of air battles in Yugoslavia. The maximum launch range of NATO aircraft was in the region of 60 km. If the drone flies in front and fires missiles being closer to the enemy, and they are guided by the plane flying from behind, an advantage appears. The drone can concentrate on anti-missile maneuvers. As a flying radar, it is also interesting especially in the absence of an AWACS aircraft. And then it arises if he is a radar to expand the radar field, this is one scenario, if he is the far arm of the other, if he is needed to cover the third. The pilot of the lead aircraft selects the scenarios the co-pilot is not needed AI is greatly simplified, but this is certainly a future perspective. Ground control is enough at this stage

[Sancho_SP]

Wrong.

Even the poor AI of a drone will surpass a manned fighter (we are not talking about other types of aircraft), because there will be three drones where there will be only one manned aircraft. Purely economically.

The drone will not try to save itself, but will attack in the first place. Trading 1 to 1 is more than profitable.

[lvov_aleksey]

Quote: NIKNN

experts cause a certain skepticism. It is associated with the drone control functionality.

Yes, there is a problem, but it can be solved. The fact is that in terms of the amount of information, the ground network is more informed than the pilot in the cockpit. Here the question is by what means to control the means of destruction. For today, I suppose, you can direct a rocket using ground means or air (it does not matter) here and AI rules. Of course, the pilot drives the BVB, but before him .... you still need to get to the BVB alive.

you do not understand, within the framework of a global war there will be no one on earth, i.e. surgical intervention, and the air should live like the sea in order to give a response many times more

[lvov_aleksey]

Quote: Sancho_SP

Wrong.

Even the poor AI of a drone will surpass a manned fighter (we are not talking about other types of aircraft), because there will be three drones where there will be only one manned aircraft. Purely economically.

The drone will not try to save itself, but will attack in the first place. Trading 1 to 1 is more than profitable.

any mistake of a programmer or rab (EMP after a nuclear strike ends all your theories), I answered another a little higher, only a retaliatory strike, and there will be no one to live
ps we will leave like dinosaurs, and after us the Earth will show who will be.
If anyone does not know EMP - the electromagnetic impulse of a nuclear explosion is measured only as a percentage of the charge, i.e. nobody knows what will happen !!!!!!!!!!! 1

[lvov_aleksey]

for everyone - watch an old movie
https://ru.wikipedia.org/wiki/%D0%92%D0%BE%D0%B5%D0%BD%D0%BD%D1%8B%D0%B5_%D0%B8%D0%B3%D1%80%D1%8B
where even AI realized that it was impossible to win

[Gagarin77]

Bredyatina. What does iron have to do with it? Any video set-top box now has a performance sufficient for everything. It's about software and algorithms. Flying in the air is even easier than driving on the road. Fewer obstacles and undesirable factors. Less cost of error. For that, the drone can go to the midget. Pilots have been afraid to do this since the 70s. a drone at altitude does not need life support systems. It is lighter at times. Therefore, the ceiling is higher. Climb 25k and fire rockets from there. 95% of anti-aircraft missiles won't hit you because they are all with aerodynamic rudders. There they lack controllability. And yours fly into dense layers from a height. This is the range ...

[Vladimir_65]

it is clearly stated: one of the directions for FUTURE use is use as an interceptor. What are we arguing about? In the future, and "on Mars, apple trees will bloom." As if right today they will be put on the protection of air lines.

[alexander ivanov_9]

Russia greatly angered Ankara by showing footage of the destruction of the Turkish Bayraktar TB2 attack unmanned aerial vehicle with its new ultra-precise weapon - the Lancet kamikaze drone. The video was published simultaneously with the information that appeared that the Russian "Lancets" are capable of successfully hitting not only land and sea targets, but also to destroy aircraft, including enemy drones.
More details at: https://avia.pro/news/rossiya-pokazala-unichtozhenie-tureckogo-bayraktar-tb2-svoim-novym-oruzhiem-razozliv-ankaru

[alexander ivanov_9]

Russia greatly angered Ankara by showing footage of the destruction of the Turkish Bayraktar TB2 attack unmanned aerial vehicle with its new ultra-precise weapon - the Lancet kamikaze drone. The video was published simultaneously with the information that appeared that the Russian "Lancets" are capable of successfully hitting not only land and sea targets, but also to destroy aircraft, including enemy drones.

[Zaurbek]

This topic is also dangerous for helicopters such as Apache, Mi28, Ka52 ... and yes, the Turks also produce kamikaze drones. The main process here is detection and identification. And the concept has been hovering for 5 years. It is also possible to place mines for tanks .... I launched drones into the air over the area and let them fly. The main difficulty is for the drone to recognize itself (friend or foe) and make the decision to attack itself. Without a person.

[Ruslan Sledkov]

AI is in which instructions-subroutines are laid on how to be in some situations. This one very quickly processes these instructions and executes the commands embedded in them. For these routines to work, you need the original data or the embedded database and the data that goes into the bean. A win in aerial combat, especially in close combat, will be given to the AI ​​who has the best instructions and who will describe the largest number of possible situations. You still cannot cover all situations. If an emergency situation arises, then it is possible to somehow teach the AI ​​to analyze the instructions already available and draw up a plan for the emergency situation. Just what is this plan? Maybe some kind of electronic nonsense? What if the AI ​​freezes at the most crucial moment? But I believe that our programmers will come up with the best instructions for the AI ​​fighter.

[Ruslan Sledkov]

It is interesting that from the phrase I wrote: "AI is an idiot, which ..." the politically correct OV removed the word idiot (written in Russian letters). Although I did not mean anything offensive to anyone, except for artificial intelligence. OV employees do something so that such situations do not arise

[Vlad Bel]

It's amazing how scribblers balance between analytics and ignorance! (about the author and sympathizers)

[Ivan90]

Experts, as always, are on top.

Nikolay Markov "Denial of loans for the construction of dreadnoughts" 1908:
"... Armadillos do not correspond to the Russian spirit, Russian character. An armadillo is primarily a machine, or, rather, a combination of many sophisticated machines. But it was not the Russian spirit that invented the car, the car was invented by the Englishman-sly. This must always be remembered. In the case of the machine, in the field of technology, in the case of military tricks, the Englishman is always ahead of the Russian daredevil, but he must never be equal to the measured Englishman with the immense scope of the Russian daredevil. Russian dare requires a light, fast, desperately bold ship, a ship that would depend as much as possible on the courage of the crew and as little as possible on machine tricks. Minosocks, submarines, light cruisers, coastal floating batteries - these are elements of the Russian coastal fleet. The Russian fleet should be coastal rather than oceanic, should be coastal, because the Russian warrior is always bigger believes in land than in the sea. All the exploits of the Russian sailors invariably occurred in view of the coast ... This is a completely logical historical fact. "A Russian sailor can do wonders of courage, can defeat anyone, but he needs to see the coast, he needs to feel that he is protecting his land."

[ppgt90]

Mother of God. What are you talking about now, my dear? It is necessary to wrap it up. Overcome Slavophilism? You would have left valiant courage to fight weeds in the country. And the most important thing! Do not take off your Panama hat, sunstroke is quite serious. Darling. Stop flirting and talking nonsense. Think over and calm down.

[Victor Tsenin]

> The reason lies in the still technical limitations of AI systems.

Here is an interesting moment, there is a gaming industry where at the moment there are extremely advanced algorithms. At the moment, you can turn your gaze, because before AI it is still like before Pluto, and algorithms are able to set the heat to many people right now)

[ppgt90]

I am more and more disappointed by the publication Voennoye Obozreniye. Well, why publish anonymous materials at the level of delusion? What can be discussed here? What useful or at least slightly interesting information can you get? Ladies and Gentlemen of the Editorial Service. You, as they say in the military, don't catch mice. Your publication will soon include posts with notes by madmen. Do we, the readers, need it?

[Evgenii Xolod]

Experts, skeptics, analysts ..... Gentlemen, all these people do not have complete information. It is owned by the customer who directly supervises the program and the developer ...