Next generation pakistan army tank to be unveiled in IDEAS2018

[JohnWick]

So where its three tank????

 

[Basel]

It's more about replacing 300odd obsolete fleet.

Yes, Al-Haider was about to add newer tanks rapidly to replace old gen tanks, AK-2 possibly be NG tank for PA in future.

 

[denel]

Regardless of what kind of tank we get we can turn Pakistan military into A.I. based warfare force using mathematically driven systems engineering implemented on off the shelf components. There are a lot of things we can do as technical ability to do so already exist within Pakistan.


For tank and other vehicle/soldier based warfare, we must develop situation aware distributed attack capability. This can be done initially using tanks only but later other vehicles and ultimately soldiers carrying anti-tank weapons. Nobody has it and we will have an upper hand always. The initiator always leads. I won’t say much but distributed attack means here simultaneously attacking vast regions by letting A.I. suggested maneuvers and targeting that maximizes damages while maintains safety margins.

Though it would mean adding more computing power in each vehicle and tank, but I have the feelings we can reduce nodal computations (e.g. each vehicle is node) to such a degree, that commercially available (and relatively cheap) data and image processing chips can be used. We will need to find a way to reduce information sharing data amount since it must be communicated in real time and in highly communication degrading battlefield environment (i.e. due to possible jamming attempts). It is possible to reduce data on some portions of the network and increase on others using A.I. driven self-organizing at-hock sensor networks such that minimum information transfer constraints are maintained or we at least have sufficiently fast processors to compensate using time-delay models.

For large scale cyber-physical systems general hardware/software verification via current logic based approaches is not possible. In plain English, mathematical proofs that all will work as required cannot be mathematically rigorously derived in general but only for specific situations). Therefore for such a cyber-physical system it would be impossible to confirm mathematically hardware-software verification, still we can definitely carry out verification with sufficient fidelity (i.e. simulation assisted models) so that even if a war last for a thousand years, we can cause massive damage with high probability, every day. Here verification means, our ability to be confident that the large scale system designed for such warfare will not suffer from unwanted or unexpected behavior, under all possible scenarios (i.e. unexpected mode of operations, glitches, data clogging etc.).

We would still need to separately process the received data to dynamically update evolution of enemy forces and create human-machine cognition of what is going on in the battle field not just for commander but for the A.I. system as well so that can keep getting smarter. That actually can be done with existing data-based A.I. (i.e. machine learning approaches).

A.I. despite its fancy name, won’t be that hard though it would take around 2 to 3 years to build basic integrated A.I. that has at one had sufficient structure to absorb ever increasing information complexity (i.e. can estimate/represent, and infer/predict dynamics of war assets, people, environmental and information based interaction of sufficiently complex nature) and yet can discriminate at a resolution where actions is most appropriate (i.e. based on the situation, and limited computational and other resources, should it focus on information disruption, feeding false information or bomb a certain vehicle cluster for maximum effect with minimum cost and so on). Current machine learning approaches can be scaled up to create such a system.


Pakistan already has people with expertise in data-driven A.I. My definition is as follows: Data driven A.I. is a kind of artificial intelligence that derives its structure and/or modifies itself in response to received data and then uses this self-modification to predict, classify/discriminate, control and or influence something out there.

Sorry... in layman world, this is still a pipe dream. I work in this field day in and out. trust me, it is decades away. we get these bubbles of breakthru. For deep learning, you cannot certifiy with mathematical proofs

 

[Dreamer.]

This is not about deep learning. Deep learning is simple basically heuristic approach. not of much use in real time varying world (i.e. nonstationary and stochastic).

End result of deep learning is just an function, portion of a hyper-surface that approximates the some sort of regularity present in the data. More complex deep learning NN structures might be able to capture non-smooth hyper surfaces but I don't think they will be robust. Time dependent Adaptation is low low since far more data is required to represent time evolution of the underlying approximation.

Most people in deep learning don't even know about these fundamental kind of things. They are just python code writers using simple machine learning work with no clue regarding even the basic underlying mathematics (i.e. nonlinear functional analysis). What they do is not A.I. That's simple function approximation. That kid of research is being done by Indian people.

We are doing far superior work in Pakistan.

So approach I was describing was NOT at all about any particular algorithm or collection of them working together (i.e. boosting). It was about something else, far more advanced and mathematically involved.

By the way If you have advanced mathematical skills, you can use use statistical learning (i.e. concept of VC dimension) to improve layers structure and to improve learning performance so it kind of gives bounds on learn-ability . Which means you can actually derive "structural stability" Theorems (for recurrent neural networks) to find the optimum numbers of layers, nodes, etc, so that over or under fitting can be avoided or optimal regularization can be imposed. It would require you to be a brilliant mathematician though. Its not the usual data learning pathway. Such approaches requires a PhD in Mathematics to say the least.

Respectfully you misunderstood what is being described here.

The proofs were not about deep learning. Proof were about cyber physical system that underlies the whole things. Its provides structure. Proofs are derived using hybrid logic, I won't describe it.

I better keep these things to myself. I derived the logical system myself. furthermore global structure of the systems is self organizing that is based on distributed dynamical systems theory. Its very complex mathematics.

The system calls in various machine learning approaches as a support for low level A.I. NOT the real A.I. The low level A.I. such as mathematically rigorous statistical learning algorithms, support vector machines, or other more general distributed intelligence algorithms work as interacting support system for basic tasks.

For some simple cases it may even call a deep learning algorithm if system proves to itself that the deep learning algorithm has done well in a certain area. All of these algorithms serve as low level machine intelligence. THEY do not provide the final A.I.

Note if we want to its possible to use nonlinear functional analysis based function approximation theory to analyses the approximation power of an A.I. algorithm or architecture including that of a Deep learning or other machine learning system by modeling the input and output as a Meta-mathematical model which can then be used to find performance bound for say a particular deep learning architecture. Such analysis is beyond most if not all machine learning community.


The kind of work you are describing is probably one of those usual using python (i.e. keras) matlab, or google, tensor flow etc. You are right You can't do much with it. That's not A.I. That's essentially glorified nonlinear regression.

I was talking about a system at a far higher level of abstraction and complexity. A self-organizing A.I. I won't disclose more. But rest assured it is very capable approach. My aim was not disclose any work but point people towards such possibles in Pakistan so that youngsters might actually learn at this level of complexity and make Pakistan better. I will off course do it with my group independently in Pakistan. Its not publish work.

Its original Pakistani research for future Warfare technology that is general and can be used to ultra-enhance any warfare system or create new ones with high degree of supremacy over others in absolute terms.

Its based on very abstract mathematics. I won't disclose for commercial reasons but some of that mathematics includes very high level abstract algebra (i.e. we extended current mathematics to define our own new algebraic structures etc). Other mathematics includes non-smooth analysis and so on.

For us its only few years away with basic system may be in 3 or 4 years or so may be earlier or bit late. Advanced mathematics shortens the pace by decades. Being Pakistanis does not mean we can't be brilliant mathematicians and or engineers. We are.

So respectfully Its not that unusual machine learning stuff. its totally different from your area of work. You are just describing ordinary machine learning algorithm namely the simplest of them of all that deep learning. Thats the simplest thing out there. Even more complex machine learning algorithms have limited capabilities. This is not about undergraduate degree level work or high school implementable deep learning algorithms.

This is far far different. TO understand this work, one has to be at a very advanced level himself with mastery in vast areas of mathematics.

I have the expertise to say that and there is nothing like out there like it. who is he layman you are referring to?. Everybody who knows some small area ends up giving a huge generalized expert opinion often since their ego is bruised.

I just realized you don't understand most of this material since you used the term "deep learning certify". There is no such thing as to certify.

The correct term was modal based verification (using say some version of modal logic logic) and its was not about deep learning. It was about cyber-physical system with A.I..


Can I ask few highly technical question to find out your level of expertise based on your decades of experience off course, before you say any things?

Just to establish that I know, you even understand what I am talking about here. Otherwise whats the point wasting my valuable time arguing over things only I understand. Its okay to say its beyond you area of expertise. OR is it?

Honesty is what defines us.

Difference between Knowing few simple machine learning algorithms and modeling systems using abstract algebras structure along with host of other mathematics is vast. Way too vast. There is NO meaningful comparison.

Respectfully it is NOT about you, it is about real rigorous mathematical science and engineering. Science is truth. It is what it is.

The kinds of things we are working on, are very abstract level representation, prediction and analysis and control. There is no meaningful comparison with any thing in the data based usual A.I. .

So what about the Tank? Was anything revealed at IDEAS? For the rest can you please open another thread.

 

[BATMAN]

@Allahistan Hello man, i will not quote your write up. i can understand what you talk about. I also agree, that you better delete your write up.
keep up the good work.. wherever you are doing it.

best regards.

 

[CriticalThought]

Thank you brother

Thank you brother. You are right I think I should delete all my technical posts. I appreciate your encouragement since I was terriblly confused as in, to what to. Specially because Once ready this kind of A.I. has applications for things such as

ability to find pattern in Pakistan (using mobile, and other data, it could log in in everything) to overtime locate spies, terrorists, and other criminals. we can for example log into into Facebook without password or decryption. or in secure communications of any kind again no decryption. As surprising as it may sound, it will actually be possible.

Even in engineering we can do thugs. for example we can ask A.I. to jointly optimize for stealth and aerodynamics with structural constraints. It will do it by first creating an exact or approximate aerodynamics model (i.e. surrogate model) for the configuration, as well as far field profile model (Radar return at all locations at given distances) and then search using these model the shape that has best aerodynamics performance yet lowest RCS. Better it can create scattering patterns by creating active currents, along with the induced currents due to incident radar waves so that it can almost exactly cancel out all or most the radar returns. at the detected enemy tracking radar locations, creating almost perfect stealth.

I am just saying all so that those teenagers who might studying, that they realize that we can do everything and at a very high level. some of them might be inspired to study advanced science engineering. I hope so.

Once again thank you. I was really confused since i wanted to delete all posts since its not right to disclose such work. Yet i was under pressure to write, including on stealth design as well as RAM synthesis.

Since many people here on forums think that Pakistan cannot design stealth aircraft as well as other nation. Those who are actually involved aerospace engineering projects in Pakistan military research facilities. they feel bad reading negative comments like this.

Since they know they have technical ability to do so. They can't themselves write as they might get arrested. So some family friends have pressuring me since last year.

anyway your encouragement to delete my tech related posts helped. It helped reduce my confusion. I have deleted them except one which I can't but I will get it done. Thank you.

Instead of dropping hints on forums, let us see some hi-tech products rolling out. The proof is in the pudding as they say.

 

[AZADPAKISTAN2009]

Hopefully not this tank

 

[Muhammad Omar]

Any update of the TANK

 

[waz]

I hope the VT4(my personal choice) is ordered soon, it won't just plug the quite large gap Pakistan has v the Indian armoured corps but also give it an edge v the T-90S.

The rest can be filled with AK 2.

 

[帅的一匹]

I hope the VT4(my personal choice) is ordered soon, it won't just plug the quite large gap Pakistan has v the Indian armoured corps but also give it an edge v the T-90S.

The rest can be filled with AK 2.

China can massively produce it and deliver it to PA asap.

 

[waz]

China can massively produce it and deliver it to PA asap.

Something needed now brother, it looks dire v the IA at this point in time. This should have taken place a while back.
I'm pretty sure the VT4 that hopefully Pakistan buys will be a more potent machine on the level of the Type 99.

 

[帅的一匹]

Something needed now brother, it looks dire v the IA at this point in time. This should have taken place a while back.
I'm pretty sure the VT4 that hopefully Pakistan buys will be a more potent machine on the level of the Type 99.

VT4 is a scaled down version of T 99A2, not a watered down version. China never cheat Pakistan in military assets. we are brothers.

PA is making the right decision at this one.

its safe to say VT 4 is the best tank technology China can offer now.

not including the new generation MBT under devlopment

 

[Signalian]

I hope the VT4(my personal choice) is ordered soon, it won't just plug the quite large gap Pakistan has v the Indian armoured corps but also give it an edge v the T-90S.

The rest can be filled with AK 2.

Food for thought...

Convert old T-59 to more AZ.. but add more IFV and APC, MLRS, Medium Alt SP AD and other support elements. effective?

 

[帅的一匹]

Food for thought...

Convert old T-59 to more AZ.. but add more IFV and APC, MLRS, Medium Alt SP AD and other support elements. effective?

China is modifying T 59 to unmanned tanks and fire support vehicle.

 

[Signalian]

China is modifying T 59 to unmanned tanks and fire support vehicle.

Good for China.
Threats to Pakistan and China are different in some aspects. This dictates a difference in tactics, strategies and doctrines also.