By Joseph Mazur

Although AI is a hot topic in the media, we hear little about its military uses. AI, machine learning, and the brain-computer interface will undoubtedly have a dramatic effect on battlefields of the future, and hence potentially on global balances of power. Are we prepared for that?

– Daniel H. Wilson ^[2]

Many years ago – never mind how many – as a graduate student at MIT, I was treated to alluring conversations about a relatively new research lab involved with machine vision, robots, and artificial intelligence. In the late 1960s, a whole department was devoted to artificial intelligence; the thinking was that machines could outsmart humans in various ways, including beating them at games like chess. Some staff and PhD students in the math department had switched departments because their instincts portended extraordinary excitement about the potential of a new research lab at MIT called MIT AI, a spin-off from what was then a small department researching operating systems, artificial intelligence, and computation theory. AI was then crude or, at most, narrow by today’s standards; indeed, some AI specialists working specific tasks call it “Narrow AI” (NAI).

The most exciting advance was a joke played by Professor Joseph Weizenbaum, who programmed a computer named ELIZA to act as if it were a human psychotherapist; it responded to users’ words by reorganizing the words to repeat them as plausible questions. Expectations were high, coming from a promising paradigm shift in believing that fundamental digital machine language, represented by sequences of 0s and 1s, could someday mimic human biochemical signals that trigger cognition. After all, in those days, computer buffs felt that everything hinged on computer models. And they were somewhat right. In that future, the MIT AI Research Lab was able to easily recruit young research staff and PhD students from the math department. I was not one of them, but I did go to a meeting with the director, Professor Marvin Minsky, who never said the word “artificial” but rather talked incessantly about what he called “intelligent machines.” I thought all machines worked intelligently but could not accept the simplest notion of an equivalence between machine and human intelligence, which must include imagination and emotions.

In 1987, we had a comprehensive understanding of how a machine can have what we think of as intelligence. That was 39 years ago, when an IBM computer known as Deep Blue beat the chess champion of the world, Garry Kasparov, in a chess match. Kasparov had that unique human skill of being able to process five moves ahead and possibly far more when needed. ^[3] For any chess match, the number of possible moves, countermoves, and outcomes is as vast as the number of stars in the universe. ^[4]

 – Garry Kasparov, Playboy interview ^[5]

In the 39 years that have passed, artificial intelligence has moved from being able to spot far more than just five chess moves ahead. Doing so for chess is far harder than beating anyone in checkers. These days, AI can go further to beat anyone in a match of Go, one of the oldest games in history. With data banks of millions of possible outcomes in games like chess and Go, AI is the master winner of all games that require 100 percent of skill, but that does not eliminate games of partial luck. For games, AI doesn’t fully know the difference between skill and luck; it feeds itself on its immense collection of data that points to all possible outcomes. 

We have been inundated by the news media’s volume of AI reporting, especially after witnessing the explosion on November 30, 2022, when OpenAI released ChatGPT, a tool to generate text, speech, and videos simply by user commands. From the opinions of political pundits to those of experts, we seem to be somewhat enlightened on the advantages and dangers. It is and will always be considered another dawn of the digital revolution, a societal future shock that brings with it bewilderment regarding what it is. We might not know what it is, but we do know it includes harmful effects of mis- and disinformation, raising the question of preparedness for such a blow against the way we build and process truthful information.

Machine engagements of conflict

What about war? Not chess, Go, or any competitive sport, but rather a sustained armed conflict causing battle-related human fatalities. All wars are deeply connected. AI could, or at least should, admit that its cyber-appendages can draw strong and elastic associations between one war and all.

The strongest connection is the unknown probability of winning. Like chess, the fundamental military model, each planned move of an armed conflict goes through a maze of possible realignment combat actions, each leading to the next. The “and then what” question arises. And just as in chess, that forward action is one of a myriad of others likely to be missed by the military commanders on one side or the other. That’s where military AI comes in: to foresee not just the ensuing realignment of combat decisions but also those in three, four, five, or 25 undisclosed possible adjustments down the line.

The most current brutal wars

Take the Ukraine-Russia war as an example. Russia failed to perceive beyond the second tier of possible consequences of its invasion. Military intelligence reports suggested that the invasion began with poor planning, readiness, and aged equipment. But had they relied on even the least sophisticated AI tools, the strategic moves would have advanced far more effectively. AI would have warned of the challenges, such as Ukraine’s military high motivation and Russia’s low combat readiness. These are warnings that AI does quite well. It would have signaled the likely finding that the West could supply Ukraine with the most sophisticated military equipment that would halt Russian air and sea operations. It would have given instructions on organization structure and even steps to avoid losing 16,071 military personnel killed in action in that first year after the invasion.

Another example is the Gaza War. After almost two years of war, we do not see a plan to end the Hamas-Israel conflict, other than an expansion of its military offensive in the Gaza Strip, codenamed Operation Gideon Chariots, aimed at defeating Hamas by destroying its military and taking control over most of Gaza. It seems clear that the Israeli Defense Forces are aiming to destroy Hamas, with a failure to see a few more possible strategies to gain any sense of what is likely to happen, say, in the third year of the war. AI would have warned the IDF about the possibility of hunger and a humanitarian catastrophe. It could have advised a plan among many that could have avoided the current explosion of international upsets. Instead, Israel has locked itself into a plan to demolish Hamas.

Wait! What and who is left of Hamas? With most of its military leaders in the Gaza Strip assassinated, what could be left to demolish? That war is essentially over. So, what is Israel’s next goal if there is no one left to kill? Perhaps it is the return of hostages. But is there a clear next move that will free a hostage or two? Or is there some secret political plan? As Ami Ayalon, an Israeli politician and former head of Shin Bet, Israel’s secret service, and commander-in-chief of the Navy, recently wrote in Foreign Affairs, “Wars without a clear political goal cannot be won. They cannot be ended. The longer the vacuum in Israel’s planning persists, the more international actors will have to come together to prevent an even worse catastrophe than the one currently unfolding.” ^[9] You see, here is the problem. With almost 75 percent of Gaza being taken by the IDF, and with over a third of Hamas fighters killed, Israel’s initial mission to defeat Hamas and bring back the living and dead hostages could be on the brink of being achieved. And yet, the war continues elusively. The IDF reported, “From the intelligence and findings on the ground, most of the Hamas Brigades have been dismantled. It is estimated that most of the battalions are at a low level of competency and can no longer function as a military framework.”  ^[10] And yet, new fighters, perhaps thousands of recent recruits, in northern Gaza have appeared to rebuild brigades. “Hamas can draw on these 2 million people for recruits. Most Gazans are young, with many under the age of 25. This means Hamas needs to recruit a few percent, and it has many forces.” As the Times of Israel put it, “Hamas returned. Or perhaps Hamas never left.” ^[11]

Israel, being one of a half-dozen leaders in AI, must have had AI intelligence flagging the likely possibility that among the 2.1 million Gazans and Palestinian refugees remaining in Gaza, there is enough of a pool of reinvigorated young recruits to reinflate the forces and battalions needed to keep the war going. On July 28, Ron Dermer, Israel’s Minister for Strategic Affairs, spoke with David Friedman, a former US Ambassador, saying, “No outside force will be able to take control of Gaza if there are still 20,000 Hamas terrorists running around the territory. No investor will rebuild Gaza if Hamas remains, and things could flare up again. This is our opportunity to put Gaza on a different track and ensure security for decades to come.” ^[12] So, did AI give Israel that information earlier to pack some reasoning into its ongoing Gaza plans? Most likely, but they must not have been taken seriously, because political and commercial plans trumped reason. The warnings about new recruits must have been considered in planning adjustments as the war continued; Hamas lost 20,000 fighters in the beginning, but in two years of the war, it gained 20,000 to refill its brigades and battalions. Filling ranks to continue the war indefinitely was surely flagged from the beginning. AI would have warned that the active military objective would keep the war going chaotically under its chosen plan because every killing angers a noncombatant young person willing to die to satisfy emotional revenge. Did anyone listen to AI, or was its plan dead on arrival? 

A sister of artificial intelligence: machine learning

While AI is, and will continue to be, used in armed conflicts to improve decision-making, machine learning (ML) contributes to enhancing autonomous weapons systems , “exacerbate existing power imbalances, and blur the lines of accountability in warfare.” ^[13] Those are two distinct but intertwined roles.

For the past decade, the thermostat in my house has been learning my habits. It’s not AI per se, but ML. There is a difference. AI reaches for data; billions of pieces associated in one way or another with other pieces that can be algorithmized. ML has a correction function, a guessing game that makes decisions based on routines that work and those that don’t. Its functioning is not unlike human decision-making if we consider that we decide on the roads to take that are best for living well. My smart thermostat is just a guessing machine that goes by my routines, nothing more. The same works for my car’s lane change assist system that alerts me to dangers and could – if turned on – offer steering assistance with the help of sensors and cameras. It is just following what it knows and sees as potential hazards. Some of the newer cars collect data to be later used for the next level of safety. So, AI and ML mingle to help each other. Each has a function.  

Here is the right question to ask: though AI can check human error and likely improve some efficiency, what could happen when its systems malfunction or become weaponized? ML could fail and turn a machine into a weapon; however, it does not make decisions – humans do. Self-driving cars control steering and braking by cameras and sensors to avoid calamities. And, in a stretched sense, they rely on physical data supported by precise satellite and road reports. Military AI is different. For that, ML is joined at the hip with AI. My article “Wars Of The Future Are Coming. Are We Ready?” reports on concerns that autonomous warfare strategies with mechanical soldiers might save lives, mostly on the side of who is using them, but how it behaves has moral and social ramifications following “the foundations of humanity itself, and who we are as a people.” ^[14]  Some good may come from AI- and ML-supported future wars of autonomous battling; however, dangers could come from those new tools of warfare that could be used by criminal players, crime syndicates, militias, and terrorists. Military theaters are constantly reviewing how far they can go with international law restrictions. 

How far can they go?

These days, almost all military operations include intelligence gathering, surveillance, pattern analysis, and analysis of enemy behavior, to optimize military strategies. NATO tells us that AI is used to identify and communicate risks or threats and give an advantage in preparing for attack. ^[15] Even though the US, China, the UK, the European Union, and countries in Africa, the Middle East, and Asia, signed a declaration in the UK (the Bletchley Park Declaration), agreeing that there are advantages, challenges, and risks, and warn that AI and ML could go wrongly in armed conflicts, they speedily continue to develop that technology with no hesitation. ^[16] The risks include autonomous fighter jets ready to attack without the human trait of instinctive hesitation. 

The photo above shows an X-62A VISTA, a modified F-16. Although it is a two-seater plane, generally, no humans are sitting in those seats during test flights. On May 2, 2024, however, Frank Kendall, former United States Secretary of the Air Force, flew in one of those seats while the X-62A entered a dogfight against a conventionally controlled F-16.  “The dueling F-16s came nearly nose to nose in a series of maneuvers within 1,000 feet of each other, according to the Associated Press, which witnessed the aerial confrontation. The Air Force hasn’t disclosed a winner.” With that success, the U.S. Air Force plans to have over 1,000 autonomous fighter jets ready by 2028. ^[17] “It’s a security risk not to have it. At this point, we must have it,” Kendall said after he landed. That brings us to a new kind of warfare.

Could AI-ML bring on more wars? AI can roughly sift through the data of two enemies to compare their military capabilities and the weight and balance of their powers. When that happens, an inevitable persuasive argument rises to aggressive heights of influence: the use of force to gain property or political advantage. AI can tell a leader whether a military success is achievable. The most noteworthy example is WWI. Although AI was not involved, since it didn’t yet exist, considerable data showed the balance of military power to be on the side of Germany. With that advantage, the Kaiser was able to play his cards of force. That force mistakenly drummed up his country’s capabilities to battle. AI, now with all its data-sifting analysis, could easily make the same mistake, for war has hidden complexities that ride on the view that war is biologically necessary for some animal species that struggle with subsistence and fight for existence. No surprise: leaders with high military power, especially those with grandiose self-image, instinctually weigh their odds in decision-making and conflict-planning. Military advantages tend to be triumphalist itches for wars. The military historian and Prussian general Friedrich Adams Julius von Bernhardi, who was “the first German to ride through the Arc de Triomphe when the Germans entered Paris”, wrote in his best-selling 1911 book Germany and the Next War, “[War] is a biological necessity … the natural law, upon which all the laws of Nature rest, the law of the struggle for existence.” ^[18] That primordial impulse will always be with us, along with the good, bad, and acceptable human urges that brought us to this stage of existence.

Delegating lethality decisions to machines

How does AI tweak the advantages or balances of power, if it does? The Vietnam War has an answer: the American strategy was based on chess, while the Viet Cong and North Vietnamese Army, under a cultural difference, played their guerrilla war as if it were a game of Go. They are very different strategies! In chess, we take out the most valuable pieces. In Go, we encompass territory. But the strongest tweak comes from AI’s decision-making after analyzing data that could be true, false, or weak. When AI sifts through its data, it can fact-check at fantastic speeds, easily corroborating a million sides, if there are that many.  When it comes to a final decision without human oversight – because, in war moves, the mass of possible actions, procedures, and episodes could be too enormous without a capacity for help from someone like Garry Kasparov – potential mistakes leading to accidental escalation are likely. That is because AI algorithms, with all their speed and capacity, can also generate realistic but fake information and shoddy, junk, or manipulated science using their own tools to flood the web, and be confused by their own creations and how they interpret them. Remember that a half-billion pieces of disinformation come from AI sifting through the world of data that it hopes to analyze without human reliance, and thereby without regard for even the smallest range of human values. But that is only one part of military AI. Another is weapon automation, and all the speed-of-light tools of administering robotic command and control attack networks in black-box deliberation and unpredictable decision-making, soon after compiling neutral and biased data, whether for preemptively defensive or offensive conclusions.  

Military AI is used for planning warfare strategies that save lives, and autonomous weaponry that sidesteps moral and social ramifications. Future wars could be less lethal if wars are to become robot against robot. How they advance regarding killing will depend on the brilliance of human strategic planning, not on AI decision-making, but rather on the continued sales of new tools used by criminal players, crime syndicates, militias, and terrorists. ^[19] So much can go wrong with weapons tied to AI autonomy that have neither fear nor emotions. Of course, military planners understand that there are natural biases that need overseeing with hopes that commitments sensibly follow international law restrictions before funding. Lives could be saved, and machines that fight without care can replace soldiers. War would then become a platform of entertainment. However, war is never simple. Moral codes of robot specs depend on how tactical intelligence is programmed regarding robot sensors that could clash with objectives, Those conflicting situations ignoring the human brain, as Daniel Wilson said, “might go berserk not knowing when to click the off switch.”

AI moves so fast that sometimes it is so far ahead of its game that it does not find rare but essential hidden nuances of change. It scans data to learn immense amounts of information, while uncovering intelligence treasures by sifting through garbage. It cannot learn much when human ideas swiftly surface as paradigm shifts, vis-à-vis the recent switch from trench warfare using tanks and armored personnel carriers to the simplicities of automated drone wars. The Ukrainian military did not invent the drone; the overwhelming drone strategy has changed the war’s dynamic to maintain an edge by drones, causing more than 80 percent of Russian frontline casualties and the destruction of almost 90 percent of Russian tanks and armored vehicles. The brilliance of Ukraine’s AI drone approach offsets the balance of power; Russia thought it had an enormous advantage with its tanks, missiles, armored personnel carriers, and planes, but its war plans, with all its AI planning, totally missed “Operation Spiderweb” backing Ukrainian warrior pride, morale, and its creations of cheap (less than $800) one-way lethal drones following simple algorithms that overwhelm enemy combatants in frontline foxholes and hitting high-value targets and multiple airbases deep inside Russian territory.

Time is moving from an epoch of nuclear deterrence hopes to AI’s military intelligence defenses. The might of nuclear deterrence depends on survival and retaliation power after a nuclear attack, making a first strike suicidal. Sam Winter-Levy and Nikita Lalwani, both at the Carnegie Endowment for International Peace, wrote in Foreign Affairs that if a state can use AI to pinpoint locations of nuclear submarines and missile sites, and to disable command-and-control networks, then, with a risky first strike, it would tip the balance of power to a position of absolute dominance. ^[20] Of course, any first strike, even an action that hints at such a blow, would be dangerously boosting an arms race.

Brain-computer interface

According to RAND, a nonprofit, nonpartisan research organization, the U.S. Department of Defense (rebranded as Department of War ^[21]) is developing brain-computer-interface (BCI) technologies for humans with neural implants to have cognitive and AI data exchanges with a computer. BCI development began as research for people with disabilities, enabling prosthetic limbs, voice recognition, and many other hopeful neurological benefits through practical means. The technology started at the dawn of the 20th century with the invention of the electroencephalogram (EEG). By the early 70s, cortical activity through BCI, including visual control of computer cursors, was rapidly experimented with human brain implant units that interacted with digital devices.

BCI neurological advances are helping people with disabilities. Now, though, the military is seeding ideas centered on human-machine decision-making in combat scenarios and military planning and tactics. After all, the military’s mission is to win wars. That harbors a frightening question: will neurotechnology advance to a level that permits combatants in conflict to ethically control weapons by thought, without fear, anxiety, or other emotions that follow a more efficient mission through control of behavior?

What?! Could a special forces unit, using BCI techniques, send and receive thoughts to and from unit commanders, enabling real-time, rapid response to threats? Yes! Dr. Al Emondi, program manager in DARPA’s Biological Technologies Office, ^[22] said, “Smart systems will significantly impact how our troops operate in the future, and now is the time to be thinking about what human-machine teaming will actually look like, and how it could offer what it needs to accomplish. If we put the best scientists on this problem, we will disrupt current neural interface approaches and open the door to practical, high-performance interfaces.” ^[23] DARPA says the U.S. military will be ready for BCI by 2050, a relatively long time from now, so there is time for human rights agencies to step up their duty to safeguard the human dignity that incorporates emotions, thinking, imagining, and the perception of reason.

As Minsky is purported to have said, although I cannot find citation evidence, “the brain is a meat machine.” Following that metaphor, and the significant advances over the last 50 years, I don’t see any evidence of that, though I do worry about how long it might be preserved.

About the Author

Joseph Mazur is an Emeritus Professor of Mathematics at Emerson College’s Marlboro Institute for Liberal Arts & Interdisciplinary Studies. He is a recipient of fellowships from the Guggenheim, Bogliasco, and Rockefeller Foundations, and the author of eight acclaimed popular nonfiction books. His latest book is The Clock Mirage: Our Myth of Measured Time (Yale).

Follow his World Financial Review column at https://worldfinancialreview.com/category/columns/understanding-war/. More information about him is at https://www.josephmazur.com/

Notes

^[1] Marwala Tshilidzi. “Militarization of AI Has Severe Implications for Global Security and Warfare,” United Nations University, UNU Centre, 2023-07-24, https://unu.edu/article/militarization-ai-has-severe-implications-global-security-and-warfare.

^[2] Daniel H. Wilson, How to Survive a Robot Uprising: Tips on Defending Yourself Against the Coming Rebellion (New York: Bloomsbury, 2005) 14.

^[3] https://www.thenewatlantis.com/publications/can-chess-survive-artificial-intelligence

^[4] https://www.sciencefocus.com/future-technology/ai-has-dominated-chess-for-25-years-but-now-it-wants-to-lose

^[5] https://www.playboy.com/magazine/articles/1989/11/playboy-interview-garry-kasparov

^[6] https://hai.stanford.edu/ai-index

^[7] https://www.tortoisemedia.com/data/global-ai

^[8] https://hai.stanford.edu/ai-index

^[9] https://www.foreignaffairs.com/israel/israel-fighting-war-it-cannot-win

^[10] https://www.timesofisrael.com/liveblog_entry/idf-rejects-cnn-claim-many-hamas-battalions-remain-operational-says-most-dismantled/#:~:text=The%20investigation%20by%20CNN%20found,and%20financially%20support%20our%20work.&text=The%20Times%20of%20Israel%20Community.

^[11] Ibid. Times of Israel

^[12] https://www.jpost.com/israel-news/article-862638

^[13] https://nrdc-ita.nato.int/newsroom/insights/navigating-the-ai-battlefield-opportunities–challenges–and-ethical-frontiers-in-modern-warfare

^[14] https://worldfinancialreview.com/wars-of-the-future-are-coming-are-we-ready-foretelling-technological-and-strategic-evolution-of-battlefields/

^[15] https://nrdc-ita.nato.int/newsroom/insights/navigating-the-ai-battlefield-opportunities–challenges–and-ethical-frontiers-in-modern-warfare

^[16] https://www.gov.uk/government/news/countries-agree-to-safe-and-responsible-development-of-frontier-ai-in-landmark-bletchley-declaration

^[17] https://apnews.com/article/artificial-intelligence-fighter-jets-air-force-6a1100c96a73ca9b7f41cbd6a2753fda#

^[18] Barbara W. Tuchman, The Guns of August (Toronto: Presidio Press, 2004) p. 12-13.

^[19] https://worldfinancialreview.com/wars-of-the-future-are-coming-are-we-ready-foretelling-technological-and-strategic-evolution-of-battlefields/

^[20] San Wubter-Levy and Nikita Lalwani, “The End of Mutual Assured Destruction,” Foreign Affairs, (August 7, 2025). https://www.foreignaffairs.com/united-states/artificial-intelligence-end-mutual-assured-destruction?check_logged_in=1

^[21] Mr. Hegseth, the U.S. Secretary of Defense said. “As the president has said, we’re not just defense, we’re offense.” “https://www.nytimes.com/2025/09/05/us/politics/trump-war-department-defense-history.html

^[22] DARPA is the U.S. Department of Defense agency responsible for fostering revolutionary technologies for national security.

^[23] https://www.darpa.mil/news/2018/nonsurgical-neural-interfaces