US plans to use artificial intelligence created for playing chess in 6th generation fighter

In the United States, work continues on the Next Generation Air Dominance (NGAD) project, which will create a sixth generation fighter for the US Air Force. Most likely, the new combat vehicle will be controlled by two pilots, one of whom is a human, and the other is an artificial intelligence (AI) system.

Will Roper, Assistant Secretary of the United States Air Force, shared this information.

The military official made a statement that the 6th generation NGAD fighter plans to use the artificial intelligence ARTUµ, created for playing chess. However, he notes that the role of AI will be auxiliary. The main functions will be performed by a person, and the electronic device will take over the duties of the co-pilot. Roper believes that AI in its actions is guided by a logic that does not always obey real combat actions. Therefore, a person will cope with non-standard situations better than a machine.

Commenting on Roper's words, Popular Mechanics writes that in the not-too-distant future, US Air Force sixth-generation fighter pilots will fly to fight together, like Star Wars heroes Luke Skywalker and his R2-D2 droid.

To implement the plans for the full use of AI systems on new-generation fighters, new communication technologies are needed, which will allow hundreds of times to increase the speed of data exchange and the response of avionics and aircraft "mechanics" to emergency situations.

Comments

[rocket757]

Those. they want to harness the strategist.
So it is understandable, you will have to teach him a lot.
If there is a case, then we will look at such a "fighter".

[LiSiCyn]

Quote: rocket757

Those. strategist want to harness

2nd pilot function ... The 2nd, as a rule, duplicates and helps the 1st. In any situation. What happens if opinions are divided?

[rocket757]

Choosing a priority, dangerous target.
It is difficult, but possible, if the correct recognition algorithm is specified.
In general, what you need is a serious, productive machine, a high-order tactical computer.

[LiSiCyn]

Quote: rocket757

Choosing a priority, dangerous target.

Not only. Reaction to enemy attack. The pilot pulled the handle to the left, the AI ​​to the right ...
An attack on a target, in difficult conditions (mountains, low altitude), the pilot takes risks, goes to the target, and the AI ​​scores and begins to deduce, substituting the belly.

[rocket757]

Obviously, "knocking down" a truly effective combat unit out of a human and AI is not a quick process.
Boom to see.

[Eragon]

US plans to use artificial intelligence created for playing chess in 6th generation fighter

This is how I see it, just like in a joke: "Two AI6s meet on a collision course." Shouldn't we stir up the game of chess, dear? "

[Cat]

you need a serious, productive machine, a high-order tactical computer

A serious system for collecting intelligence and analyzing the tactical situation does not have to be pushed onto every combat aircraft; it is enough to provide communication with an air or ground command post

[rocket757]

Communication lines, the process of exchanging information, is a very vulnerable spot in the tactics of using remotely controlled devices.
You can do that, but then duplication and other troubles are required!
Why then AI on the object itself is needed ???

[Cat]

Why then AI on the object itself is needed ???

If we leave out the fiction of the very concept of AI, then, probably, to compensate for pilot errors in control and rid him of routine actions, for example, on deck ships and VTOL aircraft. Personally, I am very skeptical about so-called "tactical" computers.
And you can also use it as a detachment and a commissioner in one bottle, which are always with you

[rocket757]

It's too early to talk about the horns, and there’s nothing.
Time will pass, and then the topic became clear. So we will discuss.

[Grigory_45]

Quote: LiSiCyn

Quote: rocket757

Those. strategist want to harness

2nd pilot function ... The 2nd, as a rule, duplicates and helps the 1st. In any situation. What happens if opinions are divided?

it is natural that human actions are in priority. AI is still an assistant, prompter, observer, navigator, etc. "Bring it, give it, go to your place and don't bother." Now, if a person performs inappropriate actions or, on the contrary, does not perform any actions in a dangerous situation (lost consciousness, wounded), then the AI ​​can take control.

[LiSiCyn]

Quote: Gregory_45

Now, if a person commits inappropriate actions

I have already given the options ...

Quote: LiSiCyn

An attack on a target, in difficult conditions (mountains, low altitude), the pilot takes risks, goes to the target, and the AI ​​scores and begins to deduce, substituting the belly.

[Grigory_45]

Quote: LiSiCyn

I have already given the options ...
Quote: LiSiCyn
An attack on a target, in difficult conditions (mountains, low altitude), the pilot takes risks, goes to the target, and the AI ​​scores and begins to deduce, substituting the belly

you did not understand a little. Examples:
1. the pilot approaches the landing, not maintaining the glide path and speed, the probability of an accident is very high.
2.An obstacle (a mountain, for example) along the course of the aircraft, the AI ​​gives a signal, but the pilot does not react and continues to drive the car towards a collision, the probability of a catastrophe, if the flight parameters are not changed right now - 100%
etc.
If the pilot is alive, controls the plane, takes risks, and the risk does not border on the probability of 100% ditching the car, the AI ​​should not interfere (except for issuing warnings and recommendations, as does RITA)

[LiSiCyn]

Quote: Gregory_45

an obstacle (a mountain, for example) in the direction of the aircraft, the AI ​​gives a signal, but the pilot does not react and continues to drive the car towards a collision, the probability of a catastrophe if the flight parameters are not changed right now is 100%
etc.

So I brought ...
A case from one's life...
I have a good friend, a Mi-24 aircraft, an Afghan. So he said that one day the "crocodile" and the crew (pilot and pilot) returned from the mission after landing, pulled out the navigator and began to beat him. They beat me hard and none of the people who met me interfered. My question is why? The answer was this. They walked along the gorge, it narrowed. The pilot, not long ago from the Union, did not really fly in the mountains. Didn't take into account the descending streams and started to output late. The navigator, who has a second handle, didn't help, got scared, opened the lantern and wanted to jump out. The crocodile at this time was already chopping the rocks with a screw. The pilot is handsome, stretched out.
The navigator was written off to the Union.
Do you think AI can be stuffed in the face, if that ...?

[Grigory_45]

Quote: LiSiCyn

The navigator, who has a second handle, did not help, got scared, opened the lantern and wanted to jump out.

AI has no emotion. This is roughly program code.
Remember one of the tragic villages of our modern aviation, the death of pilots from the "Knights" near Cam Ranh? If the AI ​​stood on Sushki and he received a signal from the host about an imminent collision, the tragedy would have been avoided. In such situations, the AI ​​would not ask the pilot or wait for a reaction. Either he would steal the planes or bail out the pilots.

[LiSiCyn]

Quote: Gregory_45

In such situations, the AI ​​would not ask the pilot or wait for a reaction. Either he would steal the planes, or bail out the pilots.

We're talking about slightly different things. I'm talking about combat work ...
You, for help in piloting.

[Grigory_45]

Quote: LiSiCyn

Quote: Gregory_45

In such situations, the AI ​​would not ask the pilot or wait for a reaction. Either he would steal the planes, or bail out the pilots.

We're talking about slightly different things. I'm talking about combat work ...
You, for help in piloting.

One does not exclude the other.
If only about assistance in battle, then, IMHO, AI here should be very limited in independence. Act as an assistant, assistant. At least he should not have the opportunity to use weapons independently. Means of defense (electronic warfare, etc.) - yes, but not weapons. It is the exclusive prerogative of the pilot, if he is in the aircraft.

[paco.soto]

At that time the crocodile was already chopping the rocks with a screw.
Tell me how old are you ??
I ask after reading your comment.

[LiSiCyn]

Quote: paco.soto

Tell me how old are you ??

For what purpose are you interested ??
I ask after reading your comment.

[paco.soto]

For what purpose are you interested ??
I ask after reading your comment.
Nothing personal: simply "chopping over the rocks with a screw" in reality means an immediate and inevitable fall.
Nevermind. Best regards to you.

[LiSiCyn]

Quote: paco.soto

simply "chopping on the rocks with a screw" in reality means an immediate and inevitable fall.

Do you know that a helicopter rotor is made up of segments? Some number of which you can lose, but at the same time continue to be in the air ...

Quote: paco.soto

Nothing personal

Are you a helicopter pilot? On what grounds do you doubt?

Quote: LiSiCyn

I have a good friend, a Mi-24 aircraft, an Afghan.

2 terms in Afghanistan 84-86 He very rarely tells, only "over a glass of tea" and if beautifully brought to the topic.
And to you, with respect!

[paco.soto]

I apologize: if only there was a video option to reply to VO !!!
Now there is no way to be in dialogue.
Regards to you.
My comment was not aggressive, it is a pity that we are not neighbors and there is no opportunity to discuss with a glass of wine.

[LiSiCyn]

Quote: paco.soto

sorry that we are not neighbors

Who knows...?

Quote: paco.soto

Now there is no way to be in dialogue.

Write in a personal, we will talk.

[businessv]

Quote: LiSiCyn

What happens if opinions are divided?

I assume that the first will disable the second and this is not a figure of speech!

[LiSiCyn]

Quote: businessv

I assume that the first will disable the second and this is not a figure of speech!

And if, on the contrary ..?
And this, too, is not a figure of speech ... Turns off the oxygen supply or even catapults the pilot, in order to save expensive equipment. He will think that the pilot is inadequate, or if the system fails - "the pilot is killed", the car is fine ...

[vkl.47]

Maybe start with Tetris after all?

[rocket757]

No, no, the best game of all time .... "Well, pogo" !!!
The wolf so recklessly caught eggs in a basket .... well, a hare, a petty dirty trick!

[Cat]

from Tetris?

No, it's better with "chapaev". This is not chess, you have to think here

[tikhonov66]

Utopia.
- at a critical moment - the pilot and the AI ​​- trivially DO NOT UNDERSTAND each other ..
eight-))).
"...
Boeing 737 MAX: The problems here are related to the autopilot and the control system during takeoff and landing. The rest were tested, worked out and still work efficiently. But the nuances associated with the climb or descent turned out to be insufficiently correct in their program structure. The pilots, just like the testers, might not have known this, and the designers, programmers might not have taken these nuances into account. Because new approaches are not so easy to find out yet. Therefore, time is allocated for technical, operational tests. Probably, somewhere in a hurry. If this is precisely established in court, then the pilots can be paid compensation. These restrictions are usually communicated, they can be spelled out in the instructions. For example, in such and such a mode this and this cannot be done, but in another, please do it. But whether the designers themselves know these nuances is a matter of future research
..."
"...
F35: As of November 2019, the F-35 software has 873 DETECTED defects
..."
- and no one knows - how many more will be found.
Iba - there are cases when errors in software - appeared / were discovered after several YEARS of error-free operation ...

- and this is MATCH ...

[Shahno]

// F35: As of November 2019, the F-35 software has 873 DETECTED defects //
It depends on what you are talking about. If about the part in which avionics works according to the number of defects schemes, it is evil, if about systems simulating the functions of intelligence, this is a necessary stage of operation (that is, mistakes are even more taught). The question of how to make the training process less costly is already an organizational and technical point.
Ps. No matter how much you teach a child at home, only in a real situation does he grow up. It's clearer, probably ..

[Roman070280]

A person controls the flight .. AI is engaged in aiming at targets and shooting .. in fact, those where large calculations and calculations are needed .. How would it be logical ..

[Tsoy]

Go horse for a century will not see ...

A week ago, U-2 dragon Lady took off with the AI ​​on board as co-pilot. Responsible for navigation and communication. Slowly move the program forward. And this is not surprising since there is a handicap at 14 years old.

[NDR-791]

Eh, I’ll finish teaching AI and retire !!!

[Vasiliev S.]

Data exchange is not clear.
With whom, what, why.
About mechanics too, limited by overload.
Strange text.

[Vovk]

The article is not about anything!
Where is the analysis of existing AI software for the military?
What model is the AI ​​built on?
Comparison with analogues in the world?

[Svetlana]

Don't confuse news with review. And most importantly, foreman, you are on the Internet, no one owes anything.

[voyaka uh]

It is self-learning software.
A minimum of information is first loaded into memory.
Further, the software itself completes the program code and itself
loads itself a database. Without the participation of programmers.

[Note]

Yes Yes. And then it turns into a skynet, and decides that people are not needed. We watched the movies. in general, auxiliary systems have been working for a long time and are now just improving a little. That's just to the real AI as far as the moon on foot.

[Cat]

The software writes itself the program code

No software can independently change the general algorithm embedded in it without external intervention, compile and execute itself.

[Shahno]

Quote: Gato

The software writes itself the program code

No software can independently change the general algorithm embedded in it without external intervention, compile and execute itself.

Let's just say that there is a movement towards the assimilation of programming languages ​​and languages ​​"natural" ....

[Cat]

assimilation of programming languages ​​and languages ​​"natural"

These are logical tricks, nothing more ... The fundamental problem is that any system can reproduce only a simplified model of itself. Incl. and man. In any software, the architect lays down models of his brain algorithms, simplified to a primitive, and there is simply nowhere to take other software.

[Shahno]

Quote: Gato

assimilation of programming languages ​​and languages ​​"natural"

These are logical tricks, nothing more ... The fundamental problem is that any system can reproduce only a simplified model of itself. Incl. and man. In any software, the architect lays down models of his brain algorithms, simplified to a primitive, and there is simply nowhere to take other software.

I actually wrote to what. With reference to my phrase above .. "Natural" language can be considered as natural software for communication algorithms of biosystems. And as you know, modern Hebrew was modified by specific people, and not by some power from above.
Regarding the cognitive functions of the same brain, we still do not have a consensus about the logical structures there ... While everyone is in search, as it were, it is possible that it is logic that plays a serious role in neurophysiological processes (although it is not completely clear which one)

[rotmistr60]

in the not-too-distant future, US Air Force sixth-generation fighter pilots will fly and fight together like Star Wars heroes Luke Skywalker and R2-D2

In the presence of imaginative thinking you will not refuse. But probably all the same it is necessary to be more modest. The fighter of the 5th generation (F-35) has not yet been fully brought to mind, but have already taken up the 6th with AI. Now this topic will be promoted in such a way that the "allies" will begin to sign up in the queue for acquisition even at the stage of drawings.

[Yves762]

The military official made a statement that the 6th generation NGAD fighter plans to use the artificial intelligence ARTUµ, created for playing chess. However, he notes that the role of AI will be auxiliary.

~~~ ~~~

[Shahno]

// artificial intelligence created for playing chess //
Very informative. So to speak, to nothing. There are hundreds of such systems for chess, and they all differ significantly. From simple expert systems to reinforcement learning and Alpha zero ... What was used in the case of an AI pilot, which system ..

[Tsoy]

There are hundreds of such systems for chess, and they all differ significantly. From simple expert systems to reinforcement learning and Alpha zero ... What was used in the case of an AI pilot, which system ..

Artuμ is based on a gaming algorithm known as μZero, which has been used to beat human players in chess and Go ...

We used the Zero algorithm.

[Shahno]

Got it .. Something I strongly doubt that it makes sense to transfer the alphazero architecture to a combat robot, especially to a high-speed object. I don't think the same transfer learning will be successful there ..
Therefore, it should be a fundamentally different architecture and teaching methods both for an individual NN and for the complex as a whole (which is far from simulating board game participants).

[Lord of the Sith]

And how much will the budget be cut this time?))

[El Chuvachino]

In the not-too-distant future, US Air Force 2th-generation fighter pilots will fly and fight together like Star Wars heroes Luke Skywalker and his R2-DXNUMX droid.

You are late, lads.

[alexniko77]

will play chess

[A. Privalov]

... The military official made a statement that the 6th generation NGAD fighter plans to use the artificial intelligence ARTUµ, created for playing chess. However, he notes that the role of AI will be auxiliary.

Now we will see the opinions of experts who claim that a program for playing tic-tac-toe will be sufficient for the tasks at hand.

[MMX]

No. Recently, in one topic, adherents of the Su-57 convinced me that all you need to do is put the second crew member and the two-seater cockpit, and then the aircraft's capabilities will be fully revealed!
The Yankees just don't know this secret.

[voyaka uh]

This AI not only defeated human grandmasters in chess,
but also other AIs that I played with.
Moreover, he confidently defeated two world champions in go.
Go is a game a thousand times more difficult than chess. Game of mathematicians.
It has countless options. Algorithm impossible
enumeration of options stored in memory.
This AI is self-learning. First he will be put on the plane as a student.
He will not interfere, but only learn. Then, as a 2nd pilot.
Then he can lead the fighter himself.

[Note]

And he will throw the pilot off the catapult, because he decides that this is overweight and interferes in a maneuverable battle

[Dzafdet]

How much will they squeeze it in? While they have AI implemented on large machines ...

[Shahno]

Quote: Dzafdet

How much will they squeeze it in? While they have AI implemented on large machines ...

I agree. Including computing resources here is one of the critical points ...

[voyaka uh]

The software runs on a computer, a little more power than usual
personal computer. This is a new generation software - no brute force
options, there is self-learning based on copying human actions.

[iouris]

I don’t understand who is teaching whom to play chess?

[Viktor Sergeev]

This is not artificial intelligence, but just a machine that calculates the options that are embedded in it. If it is non-standard, the computer will go crazy.

[voyaka uh]

This is not true. There is no enumeration of options. Because nobody laid them
in advance. This software learns "from scratch" on the fly from "opponent" or "teacher".
For most people accustomed to traditional software, this method is difficult to grasp and accept.

[Note]

And what, according to machine logic, is the copying of human actions? This will be further enumeration of options, albeit not laid down earlier, but obtained as a result of training. After all, it is the possible actions that are initially laid down to speed up the process. High-speed performance. And if it is really based on the data obtained as a result of training, then the speed of enumeration of options loses its meaning, since it relies on the human factor of the speed of decision-making, and the waiting time for verification by human parameters. In addition, according to the mathematical model, there are endless variations of the same event stretched along the timeline. What about particulars? Or has the AI ​​mastered the Okamma Method? Please enlighten

[Vovk]

Quote: voyaka uh

This is not true. There is no enumeration of options. Because nobody laid them
in advance. This software learns "from scratch" on the fly from "opponent" or "teacher".
For most people accustomed to traditional software, this method is difficult to grasp and accept.

Only the neural network learns from scratch. But she also considers a huge number of options that lead to the right decision. And her training only works by collecting her own statistics.
Real AI is now built on a bundle of 3 pillars:
1) Expert systems (ES) - based on the database of rules;
2) Neural networks (NN) - based on our own built database for choosing the best options;
3) Modeling systems (SM) - based on modeling the behavior of objects in the form of a mathematical model.
And this could be realized only in the self-defense systems of a number of countries in Russia, this is "Perimitr".
In an airplane swing, it is possible to implement a bundle of ES and NS cut off to mines, but this is not AI.
The author did not even bother to read the original source a little and stupidly copied someone else's news ... shame on VO for posting something like that.

[Viktor Sergeev]

What you described is a confirmation of what I said: the computer does not have artificial intelligence, it just goes through the options and chooses the most suitable, provided that this is already included in its program. The fighter has a minimum of time to make a decision and can only learn from previous mistakes, that is, on the ground in a pile of scrap metal. A person (or real artificial intelligence) works on instincts and intuition, taking into account previous experience, he makes a decision instantly, without even trying to think. The human brain (subconsciousness) in a critical situation works much faster than the supercomputer itself.

[Viktor Sergeev]

Try to ask the computer the following situation: the carrier (person) is in a state of poisoning, sits in a hot bath, he simultaneously has diarrhea, nausea and a state close to loss of consciousness, he has a few seconds to make a decision (crawl out of the bath without losing consciousness, crawl to the toilet, rip out what to do first) and then you will understand how a person differs from a car.

[Note]

And the degree of freedom of decision-making when receiving large-scale data before the time of notification of a person (pilot) in order to speed up the combat process, and possible failures in decision-making algorithms based on logical machine chains that are not conducted through the correspondence of human experience and norms. As an example, when entering a given area of ​​destruction of the intended targets, multiple targets were found, the parameters of which have a more dangerous signature for the attacking forces. At the same time, some of them give out a signal of friendly forces without responding to additional requests for a digital code (friend or foe), and some of them try to use this request to penetrate the operating system of the identifier. According to the so-called security protocols. AI, these actions fall under the definition of hostile, as a result of which targets with priority are attacked, i.e. Those who are closer, faster, have more opportunities for destruction due to the available weapons on board, etc. During analysis, it was found that friendly (own) targets returning from a completed mission were hit, while some of them had damaged friend-or-foe recognition blocks as a result of the influence of the electronic warfare complex and counteraction by introducing malicious programs into the attacker's software. In addition, the enemy forces were moving in the second wave, in pursuit of the first wave of attackers, which were already identified as hostile, but the ammunition was used up for the first, "priority" wave, i.e. on their own, and strangers hitting hit most of the attackers with AI. Here is a variant of a possible scenario using AI. And this is not the worst thing. All this is not taken from the ceiling. Simulation was carried out. And besides, I very much remember the case of an AI who opened fire on friendly people at a military base, on a lightly armored platform, which had to be destroyed with the help of heavy weapons. So all these reports about the introduction of AI, which gives advantages of an almost indisputable nature, at the moment is nothing more than an attempt to force a potential adversary to follow this path, incurring exorbitant costs for the development and implementation of an extremely undeveloped topic.

[Dkuznecov]

Chess logic is based on calculating the quality of the following positions,
according to principle -
Option 1. We beat the knight (+5), and put the queen under attack (+3), with us - they beat the pawn (-1), put the rook under attack (-2). Total: + 5
Option 2. .... +7
Option 3. +4

We choose option 2.
And, of course, a few steps ahead.

Something like that.

[Cat]

Enter an additional factor, for example "fog of war" - and all logic will crumble. Calculation of quality +7 will receive a backlash plus / minus 10, and option 3 +/- 1. What will you choose?

[Dkuznecov]

But nothing. It's a self-learning system.
Will remember and move on.
I wrote HOW Chess Intelligence WORKS,
not a neural network.

[Viktor Sergeev]

And it is worth adding some unknown factor, for example, a new figure and a kaput program. And logic is certainly good, but how to take into account human actions that are not based on logic. Damn, let the AI ​​try to negotiate with a woman, freeze after a couple of minutes of conversation.

[Reserve buildbat]

Oh how ... It's time to shoot, and he thinks over E2-E4

[Gennady Fomkin]

as the heroes of the epic "Star Wars" Luke Skywalker and his droid R2-D2. forward to the early middle ages

[Izotovp]

Recently, there was a lot of hype in the press about the use of AI on Russian fighters.
Still, consider the 6th generation only on these grounds ... more PR than real difference. That's when there will be a pulse engine, a nuclear installation, an electromagnetic or electrothermochemical cannon, a laser, a radio-photonic radar and an ejection capsule-cockpit ... this is something that either already exists or is already in development ... this will be the next generation of aircraft, and not promoted modernization.