Sea Control - Sea Control 422 - Artificial Intelligence in Naval Operations with Tuneer Mukherjee

Episode Date: March 26, 2023

Links1. “Securing the maritime commons: The role of artificial intelligence in naval operation,”  Tuneer Mukherjee, Observer Research Foundation, July 16, 2018.2. Tuneer Mukherjee Twitter Feed.3. “Ethical Robots in Warfare,” Ronald C. Arkin, IEEE Technology and Society Magazine, March 16, 2009.

Hello, and welcome to the Sea Control Podcast. I'm Nathan Miller. Today, Tuneer Mukherjee joins Jared Samuelson to discuss artificial intelligence in naval warfare. Mukherjee highlights the speed of development versus the speed of regulation and the ability to program artificial intelligence to follow current international law.

SimSec is looking for a volunteer to join our technical team and support our web operations. We're looking for someone with background in WordPress implementation and support, as well as knowledge in web hosting and networking. Some knowledge of identity management and security operations is also helpful. Please reach out to content at simsec.org to share your background and discuss. I'd like to pause here to highlight our local chapters.

Whether you are in South Korea, Egypt, Singapore, France, New York, India, or the Caribbean, chances are there's a SimSec local chapter near you. You can find a full listing of local chapters and contact information on our website at simsec.org. So if you're interested, please reach out. Finally, I want to take this opportunity to recommend our partners in the SimSec podcast network, The Bilge Bumps. You can find Alex, Jamie, Drack, and a pile of Iron Brew bottles wherever you download your podcasts. And with that, Kimber's Men. You're listening to Sea Control, hosted by the Center for International Maritime Security.

Hello, shipmates, and welcome back aboard Sea Control. My guest today is Taneer Mukherjee, and we're going to be discussing his ongoing research on the applications of artificial intelligence in naval systems. So, Taneer, welcome. Could you start by telling the listeners a little bit about your background, please? Hello, Jared. Thanks for having me on the show. Really happy to be here.

Greetings to all your listeners. I am a researcher on Asian security studies. I have been a part of a research team that researched naval modernization in South Asia, specifically while I was working with the Observer Research Foundation in New Delhi as a junior fellow in their maritime policy initiative. Before that, I was a researcher at the Center for Strategic and International Studies Asia Maritime Transparency Initiative, where I was looking into disputes in the South China Sea and the East China Sea, and how artificial islands were being constructed, and how different countries were competing in the same space in between maritime disputes coming out.

I have also been researching artificial intelligence in naval operations and its application, naval combat systems as a part of my research. Well, thanks again for coming on here. I did not realize that you were associated with ORF or the naval policy piece here, but you can tell our good friend of the podcast, Abhijit Singh, hello when you return to India. But as a reminder to listeners, all opinions are our own and not reflective of any institution with which we might be otherwise associated. So I'll start with the first question here. Always the most challenging in the AI space, how are you defining artificial intelligence for the purposes of your paper? And then maybe you can tell us a little bit about how your thoughts on that have evolved

because the paper was written in 2018. It is very important at the outset to distinguish between AI-based naval combat systems and AI-based naval vehicles. The artificial intelligence itself has an evolving definition in our mind in everyday usage, and more importantly, in the military domain as well.

This is because AI does not really refer to any one thing or method or a class of solutions to a problem, I would best describe AI generally as a continually evolving, self-organized community of research activity focused on developing methods of finding solutions to problems and whose ultimate goal is to replace human cognition

with some combination of hardware and software. For the purposes of my paper, while not mutually exclusive, AI-based naval combat systems that are being developed and adopted are being adopted to augment general nautical operation that require comprehensive human cognition. Through AI-based technology,

which will require comprehensive human-machine collaboration, artificial intelligent combat systems or even navigation systems will help the naval commander in making decisions that will best equip its naval force for any one scenario. When compared to AI-based naval vehicles, this might not always have a human operator

or supervisor involved. In AI-based naval vehicles, for something which we call complete autonomy, the vessel might be engineered in a way and given a predetermined combat mission, a predetermined intelligence reconnaissance mission. And then this particular vessel will use neural networks, deep learning neural networks, machine learning algorithms, and come out with its own set of maneuvers and execute its own set of tasks in a way where it will fulfill its final mission. Sometimes I feel these machine learning algorithms can operate like a black box.

That is, while the input and output of such system will be observable, the computational process undertaken cannot be fully explained. Technical explanations may require a decision made by an AI system to be understandable, leading to a possible trade-off between an algorithm's accuracy and its explainability. To help distinguish the applications of AI, I would like to go into a few more details as to how AI is currently used in naval combat systems compared to underwater unmanned vessels. The purpose behind the application of artificial intelligence to naval operations through information management systems is to influence and augment the decision-making process of the naval commander. The practice of using digital assistance for enhanced navigability is not a new concept in naval vessels, even for understanding if the ship needs some logistical updates, if the ship needs any technological aids. But the groundbreaking aspect of the next step in naval combat systems is the ability to revolutionize command and control of an entire vessel, or even an entire fleet. The AI software would give command teams the ability to monitor real-time combat situations and suitably utilize the naval assets at their disposal. These systems combine hardware and software processes to completely transform naval maneuvers.

And the system developers who are working on these naval combat systems have attempted to mimic the process of the human brain and generate machine learning processes that analyze vast amounts of incoming data from sensors, such as satellite imagery and active sonars that are already present in today's naval vessels. And then the combination of these neural networks and deep learning algorithms are then able to present a seamless operating picture to the command team and assist them in taking decisions, enhancing their human capacity. These systems will enhance fleet operability given assets are connected to each other in today's domain of naval warfare. this command control communications computers intelligence surveillance and reconnaissance that we call c4 isr will orient a system in a way where individual naval vessels will be connected to each other irrespective of their class irrespective of the fact that they're unmanned or manned submarines frigates aircraft carriers battleships unmanned vehicles can all be equipped individually with our systems, then to augment their specific functions as well. So essentially

these are deep layers of AI that can be fitted into the whole command and control structure of naval force. Both unmanned and manned assets will be connected to each other in a way where the naval commander who's in charge of the whole operation will then be able to best utilize options that are available to them, best utilize maneuvers that are available to them. And sometimes these systems will offer maneuvers or options that the naval commander can cognitively not even think of. And that is where we come in with the enhancement. The AI-based intelligence systems will cue other systems, confirm potential threats, which then will relay back information to the officers in charge. And while these AI systems

are at work, while they have some predetermined engineering and algorithms in place, the most important bit about these systems is that they will also recognize the behavior patterns of the naval commanders in charge and learn from them in return. They will run multi-agent-based simulations and enable an end user to improve their maritime domain awareness for rapid tasking, detecting, and even tracking of non-cooperative vessels or in even a situation where they need to rescue and perform missions in aid of humanitarian relief. A good example I would like to give in this scenario is for subsurface missions. Subsurface missions at times can be very taxing on human naval commanders, especially because naval commanders involved in subsurface

missions are often engaged in dull operations that require patience, skill, navigational expertise, and above all, their ability to react to sudden changes in situations. These AI-based naval combat systems, in return, can reduce their fatigue and generally boost operational proficiency. A good example of AI-based combat systems that might come into use very soon will be the newly revamped, consolidated afloat networks and enterprise services, which the United States Navy is hoping to fit in its literal combat ships. An upgraded version of the CANES will seamlessly connect ships, submarines, shore locations, other tactical nodes, and augment naval operability and shorten action cycles against cyberattacks, protecting combat systems

at the same time modern vessels are fitted with an astonishing number of sensor suits and by analyzing the plethora of information coming from these various nodes also connected to satellites this network can rationalize information and remove anomalies and assist the human command team because at the end of the day human naval commanders are receiving huge amounts of information from their sensor suits, from their information systems, from their own staff. And it can be difficult to take decisions in warfighting or combat scenarios where the decision action cycle is very short. Another example of this AI analysis being used in the U.S. Navy for the future might completely connect

all the assets that are in place whether it be aircraft carriers or submarines or unmanned vessels. Everything can be seamlessly connected in a way where human decision makers sitting in a room in Washington

can take decisions for the entire naval fleet simultaneously and the AI-based systems will be giving them options, queuing them systems and giving them ideas as to how to best achieve their strategy. So essentially, it might take out completely the tactics that are naval tactics that are decided by humans.

But that, I think, is very far away. When it comes to unmanned naval vessels, the application of AI in unmanned naval vessels can extend from general navigational support to complete function and autonomy. And that is the difference between naval combat systems and naval vessels. The key part of naval operations right now where they are deployed are they are used in non-lethal roles alongside conventional vessels, broadly classified as unmanned underwater vehicles or UUVs and unmanned surface vehicles and USVs. Most of these vessels are more or less under direct human control and lack mission autonomy. But the way the naval forces are evolving, a subgroup of these unmanned naval vessels that enjoy a certain degree of autonomy, commonly classified as autonomous underwater vessels or underwater vehicles, and they're called AUVs or autonomous surface vessels, which are then classified as ASVs. A report by the Stockholm International Peace Research Institute says when it comes to defining autonomy in such vessels, the ability of a machine to execute a task or task without human input using interactions of computer programming with the environment, these systems perform most analytical functions without human intervention and reduce the need for human crews that have conventionally performed them. It is very important to distinguish between the different stages of autonomy that Artificially Intelligent Naval Systems are endowed with.

In systems where operators are in complete control, humans are considered in the loop. For vessels which have a certain degree of autonomy but are under human supervision, humans are considered on the loop. while for untethered vessels that function with complete mission autonomy humans are considered of the loop and i think it's very important to define this at the outset these different degrees of autonomy are especially important in discussion of the type of mission these unmanned assets are programmed to execute while the application of artificial intelligence to naval combat systems and as a nascent phase it is more established in unmanned naval vessels what i mean in this is while combat systems are still not only reliant on AI or completely autonomous,

a lot of unmanned naval vessels that are already in use have been endowed with complete autonomy where they will perform their distinct combat mission or distinct mission without any human intervention. What are the ethical debates surrounding the use of AI in naval combat missions? as these technologies are being developed uh legal and ethical issues of use autonomously capable lethal systems are beginning to create greater gain greater salience what do i mean when i talk about autonomously capable lethal systems imagine an unmanned underwater vessel or an auv or autonomous underwater vessel which is fitted with torpedoes and different sensor suits and has been endowed with targeting capabilities by itself

which essentially means that there is no human commander sitting behind the screen anywhere that will control when or where this vessel engages. Now the debate around these systems is gaining greater salience because there is a backlash to having these vessels autonomous vessels have its own targeting and combat capacity

without any human intervention like how can you give a machine the ability to take lives or engage other naval forces the question comes from a very important facet of international law especially the article 36 additional protocol one of the geneva convention which specifies that states need to undertake a legal review of new weapons to ensure their compliance with international law so any autonomous vessel that will ever be deployed in essentially needs to comply with this law so therefore even as the new generation of ai-based combat technology looms on the horizon, endowing subsystems with targeting autonomy will have to be strictly be subject to strict provisions.

Therefore, even as the new generation of AI-based combat technology looms on the horizon, endowing subsystems with targeting autonomy will have to be subject to strict provisions of international law. In this specific context that we are discussing of AI-based naval operations and AI-supported systems. These AI-supported systems with targeting abilities, ability to engage hostile forces,

ability to fire torpedoes or munitions, these systems will need to be able to continuously distinguish between military assets and civilian weapons. That will then ensure the conformity of AI-based

need for the autonomous weapon systems with international humanitarian law. And this debate has assumed global significance over the last few years, not just in the maritime space, but in air defense space as well, where a lot of unmanned aerial vessels have been endowed with such capacity to engage enemy combatants. But there's always been human commanders behind the final decision to target or engage in combat. For the purposes of our discussion, it is appropriate to use the definition that is usually favored by the International Committee of the Red Cross, which uses autonomous weapons as an umbrella term, encompassing any type of weapon with autonomy in its critical functions, meaning that a weapon or a vessel that can select, search, detect, identify, track, and attack, use force, neutralize, damage, destroy targets without any human intervention. the United Nations Convention on Certain Conventional

Weapons in Geneva has been looking into this matter and countries are having discussions as to how much lethality can we endow autonomous vessel and is that even in line with the current international law

and there has not still been any consensus on how these weapons will be regulated and the international debate is going on while simultaneously countries like russia china the united states some european countries sometimes classified sometimes unclassified are developing vessels will that will have this capacity but have still not deployed that in real time and it is quite plausible that before any international convention or international regulation or idea comes across on the regulation

of these weapon systems these weapon systems might already be in existence that they are in deployed international regulations regulators can't control what what the states want to reveal there was also a coalition of non-governmental organizations that initiated the international campaign to stop killer robots that was led by human rights watch at that point and that basically came to the point where people started discussing uh you know scenarios which we have sometimes seen in hollywood movies where machines are artificially intelligent and aware uh they have surpassed human cognition they are aware and conscious of their own and they have the ability to then wipe out the human race which at this point i feel is a very exaggerated and overly

used scenario or overly simplified scenario for some people to engage the debate in because we have still not reached that point and secondly the conversation and debate is in place already so to take any point on artificial intelligence that is extreme is improper for the general discussion what the ethicists and the ethical debates lie is why or how can an autonomous vessel that thinks for itself ever engage a human target without the human without a human commander connected to it or taking any decisions what we defined earlier as off the loop but there a scope for finding middle ground just because international law says that such weapons need not exist because of ethical dilemmas that does not mean that technology cannot program these vessels

in a way where these ethically autonomous systems can be programmed to conform to norms of international law there is a scenario that these ai waste vessels are lethally enabled have the freedom to engage targets they have the mission autonomy but they can also have responsibility advisors pre-programmed into their systems additional control can be exercised by the software developers on how targeting autonomy functions at the most rudimentary step deploying a lethal or developing a lethal unmanned vessel or giving network systems like AI combat systems, lethal capabilities would be a matter of developing simple vessel identification system that can react to distinct acoustic signatures, distinct machine imageries

or distinct parameters that we human software developers think is necessary. Because we should always understand that the end of the day, no matter how much we talk about autonomy, me how much we talk about artificial intelligence the data that is being fed into the system the algorithms that are being designed for machine learning and deep neural learning are being

developed by human software developers control can be put in place controls can be put in place instituting ethical control there's the famous American robo-ethicist Ronald

Arkin has argued while designing autonomous systems there's a basis for the development of autonomous system architectures capable of supporting ethical behavior regarding the application of lethality in war such mechanism will ensure that the design of intelligent behaviors only provide responses within a rigorously defined ethical boundary the software would code the learning algorithms for machine learning and they would have superior target discrimination capabilities that allow for the creation of techniques to permit adaption of ethical constraints

and underlying behavioral control parameters. This is especially significant when autonomous systems are distinguishing between friendly forces, adversarial targets, between civilians and combatants.

And the ideas that govern weapon systems or artificial intelligence right now is ever-evolving. So before such weapon systems are endowed with lethality or are given the freedom to choose, to target, to engage, we still have the ability to design them in a way where they conform to international law. Tanir, we've been going at this for a little bit yet here and we're sort of reaching the end of our time together here so I'm going to ask you this last question it'll be kind of a composite of the two questions here but allow you to express a few more opinions here as well as

how do you see the U.S. and China competing in the AI space at sea and then what are the potential consequences of more widespread use of AI at sea? Artificial intelligence is already in use in naval and maritime operations. They can be used for intelligence surveillance and reconnaissance missions, for mine countermeasures, for inspection and identification, for survey missions of unmanned undersea cables, for oceanography mapping, for payload delivery,

for navigation. At the same time, while these vessels are being used in non-lethal missions, they can be used for lethal missions at the same time as well. And in that space, I feel the United States and China and their naval forces are at the forefront of developing AI-based naval systems that will advance and enhance their nautical capabilities. Both have devoted considerable resource to developing an action plan for the implementation of AI-based systems in their navies and have accompanied such plans with innovative operational procedures that would support the shift to unmanned maneuvers. I remember in 2016, the Chinese Navy seized a U.S. unmanned vessel, and that demonstrated the emerging competition between the U.S. Navy and the People's Liberation Army Navy in the new generation of naval warfare involving intelligence systems. Military planners in Beijing and Washington are coming up with innovative operational concepts

and at the same time are developing new-age vessels that have the ability to completely revolutionize how naval warfare works. China displayed its unmanned undersea vessel called HSU-001 in October 2019. The United States is developing the Orca AUV, which will mimic the functions of a submarine. When we are discussing such new vessels and new technology, apart from the U.S. and China, countries like Russia, India, are also being involved, are also involving their technologies and their naval forces in developing new age naval vessels and new age combat systems.

The United States Navy specifically wants to develop and procure three types of unmanned vessels, one called Large Unmanned Surface Vehicle, NUSV, a Medium Unmanned Surface Vessel called the NUSV, and an Extra Large Unmanned Unreceived Vessel called the XLUUV. They are thinking of proposing in the financial year of 2023 somewhere around $600 million in research and development for funding these UVs and enabling technologies. While Chinese development of such vessels or such systems are somewhat shrouded in mystery, there is some idea as to how the whole picture is shaping up in China. But more importantly, rather than going into the details of their own weapon systems, thinking about a space where in the next 10 years, places like South China Sea, the East China Sea or the Taiwan Strait

that are already contested geopolitically and involve the naval forces of multiple nations intercepting each other, interacting with each other, or even sometimes engaging each other in a non-lethal way. These contests in maritime spaces might be transformed completely

where both countries or navies involved might decide to take away the human element of naval warfare from them and deploy completely unmanned vessels, sometimes autonomous, sometimes manned in the way that they are controlled by a human commander. And you might see countries engaging each other in a way where even autonomous nuclear-armed vessels are involved. there are spaces that are contested in other parts of the world especially somewhere like the indo-pacific or somewhere around the arctic where you can see countries like russia or nato

engage each other or countries like india and china engage each other somewhere in the indian ocean so as as the maritime space evolves it would be very important to see how countries develop themselves, just not the United States or China, how they endow their vessels with lethality, what kind of technology is being used by them, and also, very importantly, what kind of naval systems are then put in place. When talking about some of the challenges and consequences of widespread use of AI at sea or for naval operations, it's very important to talk about the swarming technology that has been a part of a lot of debate that has come through when it comes through AI-based

systems. Swarm attacks are essentially high-risk coordinated assaults that sometimes target numerous targets. They are extremely difficult to defend for a variety of reasons, because one of which is simple numbers and redundancy. Because you will essentially have a group of unmanned vessels that are connected to each other and that have their autonomy endowed inside the system. Essentially, navies will be giving commands or missions to these swarms of drones. So imagine the thousands of unmanned drones that are in the maritime battle space that have been given a mission. and both China and the United States are working at this point

to evolve swarming techniques to evolve how a super swarm could probably be in use where thousands of drones or thousands of vessels might function with perfect synchronization they will perform a choreographed pre-determined route these individual drones might have

awareness or not awareness of their surroundings The US has already made a lot of progress in this area. More importantly, it is essential to recognize that while all this debate is taking place, the vessels and the missions and the artificial intelligence systems, and it seems so progressive It is easy to forget how easily such systems might also be having countermeasures, especially when it comes to hacking or GPS jamming such systems. Imagine the possibility of a naval task force, UUV or USV, being hacked and used against it.

hack-proving such unmanned systems should also be a priority for developers and that will be one of the consequences that will come through that cyber security will need to be updated cyber security measures will need to be upgraded at present the majority of applications to AI naval operations are dependent on the human operator making human machine calibration extremely important naval combat systems

only perform analytical functions in combat scenarios where the final decision is still with the human commander but when it comes to developing systems that will be completely autonomous the belief in the system's

infallibility can lull operators or navies into a false sense of confidence in conclusion I would like to say that naval operations have been transformed with every era of human development and as this new age of network centric warfare begins naval planners will need to adjust and account for a full set of possibilities ai-based systems offer the opportunity to go beyond what is humanly possible

and gives maritime security missions in new zealand from coastal defense to anti-submarine warfare to even endowing such vessels with nuclear technology or nuclear weapons with completely transform how operations, tactics, strategies, and naval maneuvers work. Yet, currently, these technologies only grant a certain technological edge in the battle space when the equation is asymmetric. And that is the present-day scenario. Naval operators, planners, strategists need to sit down, think hard as to where

this road to automation will lead them. While we shan't dismiss the development of AI-based systems, there would be a huge risk and a lot of costs involved in the long term and a step up from the current regime where human commanders will be replaced by AI-enabled systems. In the future, a face-off between navies that are both commanded by AI-based systems? We cannot even envision what a scenario like this would lead to. In such a scenario of hyper war, where the decision action cycle collapses completely, the lines between naval missions, naval warfare will become blurred,

where the better technologically equipped AI system looking at its surroundings, at its external environment, learning from the technology and the human interaction will be the winner at the end of the day. Are we approaching that world? Are we approaching that space? I think only time will tell us. So at this point, it is only recommended that the control of dual-use technologies that does not hinder scientific research and civilian applications on autonomy and economic development be researched and thought upon and examined in a way that will be fitting for all dimensions of humanitarian law or naval missions.

And for also at the end of the day, in a way which gives us more security as countries and as nations, because we should all remember the technological advances and all the warfare that we are currently talking about is to guarantee us a more secure world, not a world where we are more insecure

just because AI-based technology is involved. Well, I'm sorry, that's all that we have time for today. I'd like to thank my guest, Tunir Mukherjee. Tunir, where can we find you online and what are you working on next? you can find me online on my footer account which is m-u-t-u-n-w-r mutineer i'm currently working on a very different space i am researching the civil military relations of the indian armed forces and how different parts of the armed forces and the civilian government interact with each other. I'm also developing some research into how naval modernization is coming through in South Asia, especially when it comes to new age vessels, the geopolitical nature of China's involvement in the Indian Ocean, and at the same time, how the whole concept of the

Indo-Pacific as a mental map is being developed and thought about by different parts of the world. well thank you again for joining us the paper which i i didn't mention at the top of the podcast the uh paper title but there will be a link to it in the show notes the title is securing the maritime commons the role of artificial intelligence and naval operations thank you for coming on today to the listeners thank you for tuning in we'll see you next time © BF-WATCH TV 2021 Thank you for watching.