Warfare is changing faster than our military and our military-industrial approach. The warnings of world leaders, including the Prime Minister and Chief of the General Staff, that war is imminent have had little effect on our rate of preparation or adaptation 1. Almost no one, including those working within it, thinks the speed and scale of change in the UK Defence equipment programme in the last two years are adequate. If it were, the war in Ukraine and Chinese exercises around Taiwan should have offered little cause for alarm. ‘Everything is fine’ is a proposition few would defend. Our Parliament states that we are not ready to fight a major war with our current equipment and industry approach. Deterrence exists in the minds of our enemy and ill preparedness undermines credibility; not only does it make us less likely to win, it makes war itself more likely. If few individuals are satisfied with the status quo, why as a collective have we achieved so little change? Machiavelli would have a suggestion.
“There is nothing more difficult to arrange, more doubtful of success, and more dangerous to carry through than initiating changes. The innovator makes enemies of all those who prospered under the old order, and only lukewarm support is forthcoming from those who would prosper under the new. Their support is lukewarm … partly because men are generally incredulous, never really trusting new things unless they have tested them by experience.”
Niccolo Machiavelli
Changing the system that retards Defence’s ability to adapt incurs the frictions he described centuries ago. To disrupt the dominance of those too comfortable, before the disaster of war falls, this article aims to pick a fight. Cunningham’s law states “The best way to get the right answer on the Internet is not to ask a question; it’s to post the wrong answer.” This article introduces five provocations. We hope to provide arguments for the innovators, both inside and outside Government, to use to help drive change. But even here we want disagreement, seeing your own argument played back to you can help you see its flaws. For those that disagree, we hope you’ll see that, as John Stuart Mill wrote “the nonconforming opinion is needed to supply the remainder of the truth, of which the received doctrine embodies only a part”. Help us improve our arguments with your challenge. You came here for an argument (channelling Monty Python). We hope you enjoy it.
5 Provocations – what we are doing now is not good enough; we must:
- Plan for dissimilar re-armament: What we deploy in month three of the war will not be more of what we deployed in month one.
- Change the equation: move to $ cost to $ damage model:The economics of war have changed; we must too, or we lose.
- Link frontline to factory: War is a learning competition, and we cannot afford to be in the slow class.
- Rethink the roles of air power: Particularly control of the air and attack, in an uncrewed age.
- Accept that the future is uncrewed: The role of humans in warfare, at all levels, will change much more than is generally assumed. We need a plan for uncrewed technology at scale.
Taken together, these measures increase our chances of winning a coming war, thereby making it less likely we have to fight in the first place.
Dissimilar Rearmament.
By month three of the next major war, the aircraft, ships, and tanks that we start the fight with will be reduced by attrition. But we know now that we will not be able to replace or grow the numbers of the key platforms in the current equipment programme fast enough to keep fighting. They will need replacing, and we will need them in weeks and months, not years. We will need dissimilar rearmament.
Neither side will be replacing their aircraft fast enough, but liberal democracies are far more dependent on airpower.
In contrast, one UK drone manufacturer, Callen-Lenz, developed their uncrewed system from concept to deployed capability, with production rapidly and highly scaleable, in just 3 months.
Speed of development and manufacture will be critical. We will switch to manufacturing equipment that can be built fast enough to keep the enemy at bay, or we will lose.2
Whomever the adversary might be, we must assume that they will have calculated our likely attrition rates against their defensive systems. If we are to deter war, we must change this. If we have to fight, we will have to change the economic equation of how we fight.
Changing the Equation.
We mean this literally: capability cost per $ damage inflicted ($ cost/$ damage). If destroying an SA-21 at greater than 70% probability of success takes 30 drones, or one F-35-launched Paveway-4, we can calculate the costs of the procurement and deployment of those 30 drones and contrast them with the costs of the F-35 and the launch of the Paveway-4. We can calculate the risk of losing an F-35 against the probability of loss for the drones. We can calculate the relative risks of failure; in theory, you might lose 10 or 20 of the drones, and still achieve a kill, whereas the loss of the F-35 probably means mission failure. You can also calculate how much it will cost and how long it will take to rebuild these capabilities.
If we made this central to our operating and procurement concepts, we inherently create an incentive system for defence industry to compete to produce the most financially and militarily efficient systems relative to the effect they achieve. Contrast this with the incentives in our current model. Today, defence industry responds either to: a detailed tender of requirements; or (rarely) a contract for spiral development; or some innovation objective, against a predetermined budget – largely separating considerations of the costs of equipment from its operational employment.
For example, consider the cost of aircraft. It is estimated that 1 x fast jet mission costs ~£1.3M. 1 x Tomahawk missile costs ~$2M, equivalent to ~20 of UK drone manufacturer Callen Lenz’s higher-end drones. George Shipman of Callen Lenz points out that 1000 of their high-end systems can be bought for the same cost as a single F-35. These same drones can be manufactured relatively easily at 1000 a month. A rate that can be scaled up much more with limited investment, adding scale to speed as a key advantage in the manufacture of uncrewed systems.
On the other side is $damage. An S400 system costs >$500M, equivalent to ~6,250 of our earlier referenced drones. One S400 missile launch costs ~$2M per missile (illustrative: online cost estimates vary), equivalent to ~20 drones. And an S400 takes nearly 2 years to make and deploy. In June 2024, one of Russia’s most advanced fighter aircraft the Su-57, estimated to cost between $35-54M per aircraft was destroyed by two one-way attack drones. Calculating the $cost/$damage ratio based on our figures for Callen-Lenz’s drones (we have no information on which drones were involved), we can see the cost-to-damage ratio ranges from 0.148% to 0.229%. Russia is believed to have only 15-20 of these fighters, to be struggling to meet production targets, delivering zero in 2024. RUSI’s Justin Bronk notes, this loss represents a 5% cut in the frontline fleet of Russia’s most advanced fighter aircraft, and “Taken together [with other measures], Kyiv appears to be pursuing a clear strategy to force the VKS to either vacate its bases within several hundred miles of Ukraine’s borders or dedicate an inordinate quantity of air defence systems to defending them.” The economics for many of both our enemies’ high-end systems and our own no longer work.
The changing dynamics of defence economics are well illustrated by one recent estimate from Ukraine. Wired magazine reported one Ukrainian drone manufacturer comparing the costs of killing a tank at medium range with artillery – which generally takes 4-5 artillery shells at a cost of $8000 per shell, vs 2 of his drones, costing between $400 and $2,500. Estimates for a Russian tank costs vary, but one unclassified source suggests a T-90 tank costs $2.5M. To make the economics clear:
Drones cost per kill = $5000 / $2,500,000 damage = $.002 cost per $ damage = 0.2%;
Artillery shells costs per kill = $40,000 / $2,500,000 damage = $0.016 cost per $ damage = 1.6%.
Another way this can be said is that costs per kill are 700% more expensive for artillery over drones. But even this is an incomplete model, it only includes the munitions. How much was the launcher for the shell or drone? What cost the logistic chain to deliver launcher & munition to the launch point?
Independent of our later provocations, this specific provocation is not advocating that F-35s or Tanks or Artillery or UAVs are better as a preferred type. The provocation is about adopting a system that is ruthlessly rational about how such decisions are made.
This is not a new lesson: In March 2017 – seven years ago – US Army General David Perkins revealed a US ally had used a $3 million Patriot missile against a quadcopter that cost $200 from Amazon; Perkins stated: “I’m not sure that’s a good economic exchange ratio.” We must change the equation as a central component not just of procurement, but of our theory of winning.
Frontline to Factory.
Industrial capacity is critical to deterring war and to winning it. Yet liberal democracies have let their defence industrial base atrophy. Many still struggle to harness their technology companies for national security and defence. Most face major demographic challenges. At the outbreak of war, with global supply chains severed or under threat, and rapidly expanding the military becomes essential, there will be demands to move workers from our largely service industry-based economy to the farm, factory, and frontline. All will be key wartime areas. Getting the balance right will involve acute and risky trade-offs. Yet if we fight as we do today, and our economy is as labour-dependent as in previous major wars, a shortage in any of those key areas could mean defeat. We must assume our potential adversaries have calculated this. To deter, we must convince them we can produce and deploy sufficient force to fight and win. Technology can help change the dynamic. Today, more than ever before, automation, robotics, and artificial intelligence can reduce a nation’s dependence on people to manufacture, provide services, and scale armed forces.
This is not just a labour shortage mitigation. The tempo of military evolution makes a much tighter feedback loop between combatants and Defence-Industry essential. A necessity that will affect not just equipment design but all the Defence lines of development. Nick Reynolds, Research Fellow at RUSI makes the point well
“The technology is moving very quickly. I would say maybe a six-week learning cycle on the battlefield…Our procurement systems are not optimised for this.”
The current war, and the next, will be wars between innovation systems.
Innovation itself – the ability to creatively solve problems, ideate, iterate, develop and deploy new equipment and new systems – not just the capabilities a nation starts the war with, will be a key determinant of victory. We need to link the frontline to the factory. No sooner will a new capability, for example a tank, an aircraft, a ship, a missile, be deployed than the enemy will begin building a counter. Lessons learned on the front – perhaps even from aircraft, ships and soldiers’ sensors – must inform iterative redesign and rewriting of both hardware design and software code alike. We will need to link the information requirements from industry, academia and others into the (automated) intelligence collection plan and process of Defence HQs and the tactical frontline. Factories will need to produce version 1.1. of a product within weeks, days, maybe even hours of the deployment of version 1. Linking the frontline to the factory and changing the equation may be the only way to regain the key advantages of speed and scale in how we fight.
Rethinking Airpower.
Air Forces must rethink air power and the division of responsibilities with the other services. There are four roles of airpower: (1) control of the air; (2) attack; (3) intelligence, surveillance and reconnaissance (ISR); and (4) air.
Today, control of the air is achieved by combat air patrols, fighter sweep, fighter escort, and from ground-based air defences, are all challenged by drones – which:
- are a low-cost way to kill the current expensive systems;
- being cheaper and available in larger numbers, can be close to ubiquitous – no nation today has enough advanced fighter aircraft to achieve air supremacy over a large area;
- are able to rapidly concentrate in time and space for sweep or escort;
- could control the air at 0-10,000ft rendering control of the air above this altitude by conventional airframes redundant, or at least of reduced utility;
- are potentially more manoeuvrable, able to react faster than crewed platforms, able to execute a Combined Air Operation coordinating more effectively, and may be better than current systems at executing control of the air as we currently do, just uncrewed.
Ubiquitous surveillance from thousands of drones and persistent surveillance from both air and space assets, combined with precision fires, have made concentrating force and achieving surprise difficult. Coordination of disparate effects and rapid and late assembly of forces will be needed to gain the initiative for surprise and concentration of force. Drones provide an ideal solution. The need for close air support from more expensive, larger assets, such as attack helicopters and ground attack fighters, may reduce – especially once lower costs drones begin targeting such assets.
The role of attack, further divided into close air support, interdiction, and strategic bombing, can and will be similarly changed by the advent of low-cost drones capable of all these roles at extended ranges. ISR will be uncrewed (and largely already is). Air mobility, from transport aircraft (dozens in development, e.g. HH-100 in China, Natilus in the US, Droneliner in the UK) for intra-theatre lift via operational and tactical resupply to last-mile logistics will be increasingly uncrewed. Much of this is possible now. The principal limits are regulatory, bureaucratic and cultural, not technological.
Ground Forces on the Front Line will be capable of interdiction in the deep and maybe even strategic strikes with increasingly long-range, precision drones, missiles and munitions. This increases flexibility, but also coordination costs. Air Forces’ justification for controlling air power capabilities remain sound – to concentrate force, prevent penny packeting, and the use of air power in too tactical a role – as airborne artillery or tanks. But the RAF will have a reduced monopoly on control of air power in future warfare. It will need rethinking.
The Future is Uncrewed.
Yet many militaries are still betting on crewed aircraft. This is an increasingly unwise wager. On 4 May 2024, EpiSci, PhysicsAI, and ShieldAI reached a new milestone in their bid to automate fighter aircraft. In just three years, they have gone from competing in simulations vs human pilots in a DARPA programme ACE (Air Combat Evolution), to flying the Secretary of State for the Air Force in an F-16 “Vista” piloted by AI, going through high-speed, high-G, dogfighting (basic fighter manoeuvres, BFM, in the jargon). In June 2024, Lockheed Martin’s Skunkworks successfully demonstrated it’s AI piloted L-29 in the air-to-air role, performing ‘standard head-to-head fights, off-aspect encounters, missile support, and missile defeat situations’ with the system continuing to exceed expectations. This should not be a surprise. For the more likely forms of air combat, beyond visual range (BVR) missile launches, the pilot’s role is far more minimal than for BFM, essentially looking at a radar scope and launching when the system identifies enemy aircraft in range, and the pilot judges the optimal moment to release weapons and escape. As one wit put it, the pilot in BVR engagements is just a meat-lever.
Narrow Artificial Intelligence, that is AI that outperforms humans in specific tasks rather than in all tasks, is more than enough to make pilots redundant, and month-by-month this being proven so. We should ask why these AI breakthroughs in air-to-air combat aren’t happening in the UK. We believe the answer is continued, misplaced and now demonstrably incorrect, AI-scepticism. As recently as 2021, a genuinely bold experiment from the Royal Air Force in imagining the future through science fiction, acknowledged the possibility of an uncrewed future but hedged with a description of ‘Olivia’, a frontline pilot, operating in 2040. The RAF’s Future Combat Air System, which will in the planned best case be fully operational in 2040, is to be optionally crewed. We think both these forecasts are likely incorrect, and that the widespread scepticism they embody is holding us back. The planning assumptions need to change.
Narrow-AI is highly likely to provide sufficient capability for an uncrewed air force fleet, perhaps likewise surface and subsea fleets, and high volume low altitude UAV systems for ground forces. But the possibility of AGI would make this a certainty. How likely is that? The UK Ministry of Defence’s best practice for forecasting the future requires attribution of probability assessments to uncertain future outcomes. The best way of deriving these forecasts is from crowd sourced judgements and those of ‘superforecasters’. Today, these forecasts anticipate:
- January 2027: Weak artificial general intelligence (AGI), one performing at the 75th percentile of human cognitive capabilities;
- February 2028: Oracle AGI, matching or exceeding all humans in cognitive capabilities, is currently predicted to arrive 13 months later;
- August 2029: Artificial Superintelligence (ASI), one that surpasses us in all cognitive tasks, is forecast to arrive within 2 years of the weak AGI breakthrough,
- 2032: Strong AGI, AGI + robotics that match us in all tasks including those needing human performance in locomotion and dexterity.
You can of course reject forecasting best practice and come up with your own prediction. Or pick one from an expert who disagrees with the above. But you should be just as rigorous: what, specifically, are you saying AI won’t be able to do, and in what time frame, with what probability? Whichever way we trend, we are making a bet on the future that directs our resources in a way that makes war more or less likely to occur, and more or less likely to be won or lost. Lazy hedging based on unspecified scepticism is a poor way to make such crucial decisions.
In the UK, we are betting on an optionally-crewed fighter aircraft in 2040, while forecasting best-practice estimates AGI in 2027/8. And these forecasts for the arrival of all four forms of AGI are trending closer to the present day, not receding into the distant future, an indication, at the very least, of the speed of advance in the technology.
The utility of crewed aircraft – and crewed ground and naval systems – is diminishing faster than our militaries are changing. This is a transformation gap we cannot afford. How do we change faster?
The US leadership in uncrewed systems owes more to political than military leadership. The US military’s reluctance to experiment with and adopt uncrewed systems was only overcome by arbitrary rule imposed externally: on 8 February 2000, Senator John Warner, Chair of the Senate Armed Services Committee, introduced a bill mandating that the Pentagon’s budget must be spent to ensure that by 2010, one-third of all aircraft designed to attack behind enemy lines be uncrewed, and by 2015, one third of all ground combat vehicles should be driverless. It was a bill that forced adoption at scale on a reluctant military, even if its provisions were later reduced in the face of Service pressure.
Beyond direct political intervention to drive change, two ways suggest themselves: introduce a Roper-rule and ‘put your money where your strategy is’.
In the US Next Generation Air Dominance (NGAD) programme, Will Roper, former USAF Deputy Chief of Staff for Acquisition, not only introduced software-defined engineering, but also mandated that no aircraft should take longer than 5-years to get from design to in-service. This was to keep up with the pace of technological change. The McKinsey Global Institute estimates that compared to the Industrial Revolution of the late 18th and early 19th Century, change is now ten times faster, at three-hundred times the scale, with three-thousand times the impact. It was also to reduce the some of the causes of defence inflation, avoiding the situations where systems specified many years earlier are overtaken by tech change, requiring costly upgrade. A Roper Rule might start with that recently announced in the UK: 3 years for digital capabilities and 5 years for hardware (equipment). But innovation must accelerate if we are to deter future wars, or win them. Consequently, perhaps a Roper Rule should be progressively reduced, to force faster change, and leave us better placed to make the step change we need in the event of war.
The biggest barrier to defence modernisation is the conservative inertia of existing militaries. One way to over come this would be to force the Ministry of Defence to put its money where its strategy is, consciously divesting a minimum of 6% of assets annually and reinvesting money into new capabilities.
This would mean a line in the Defence budget allocated for the rapid and continuous procurement of new equipment following innovation and technological breakthroughs. It would amount to about 30% of the total budget over five years and ensure that all equipment was replaced every 17 years.
An arbitrary rule would force the Services to divest themselves of kit with no recourse to inter-service wrangling. Their strategy would have to adapt to one of continuous innovation.
It is a proven approach. A McKinsey study of 1,600 diverse companies tracked over 15 years showed that those that boldly reallocated resources to new capabilities significantly outperformed others with more timid approaches, earning on average 30% higher returns. The best approach was an arbitrary rule-based approach to reallocating funds.
As Michael Palin reminds us: ‘An argument’s not the same as a contradiction. An argument’s a collective series of statements to establish a definite proposition.’ Our five provocations are the collective series of statements. Our definite proposition is that: our current plans are not fit preparation for the next war. Change our minds.
Keith Dear
Dr. Keith Dear is former a Royal Air Force Intelligence Officer and Expert Advisor to the Prime Minister, now leading emerging technology development in business. He holds a Doctorate in Experimental Psychology from the University of Oxford, an Exec-MBA from the University of Cambridge, and an MA from King’s College London. He writes here in a personal capacity.
Al Brown
Al Brown is an Associate Fellow at RUSI, a former Army Officer, a former Visiting Fellow at the University of Oxford, former adviser to the UN on AI and legislation, and now director at a neurosymbolic AI programme. He writes here in a personal capacity.
-
Al Brown#molongui-disabled-link
-
Al Brown#molongui-disabled-link
-
Al Brown#molongui-disabled-link
-
Al Brown#molongui-disabled-link
Footnotes
- ‘This Is Our 1937 moment,’ U.K.’s New Army Chief Says of War in Ukraine – The New York Times (nytimes.com) Two years on from this stark announcement nothing comparable to the changes of 1938 and 1939 have occurred. By 1939 the Schedule of Reserved Occupations – those vital to running the economy – had been completed and published. The foundations for the Home-Chain system of radars, Spitfires and Hurricanes had been laid. In 1937 the Air Raid Wardens were created. Within two years 1.5 million volunteers registered. Imagine Gen Sanders speech at RUSI in 2022 creating 1.5m people ready and organised to help by 2024.
- Source: [Norman] Augustine’s Laws (via Wikipedia) and Davies, N., Eager, A., Maier, M. and Penfold, L., 2011. Intergenerational equipment cost escalation. DASA-DESA Economic Working Paper Series, 1, p.32.