Armed forces are turning to generic technologies, like AI 

Eight months since the begin­ning of the conflict in Ukraine, many les­sons have been lear­ned from this conflict, nota­bly the return to high inten­si­ty. One of the major issues ari­sing from this, which goes beyond the mili­ta­ry field, is the ques­tion of inte­gra­ting new tech­no­lo­gies to pro­vide an ope­ra­tio­nal advan­tage for the armed forces. 

Part­ner­ships, sup­por­ted by the US Depart­ment of Defense, with Micro­soft or Ama­zon Web Ser­vices (AWS) to host Ukrai­nian data that could be sen­si­tive to Rus­sian cyber-attacks, or the use of Elon Musk’s Star­link mobile ter­mi­nals to connect to the Inter­net via satel­lite, have high­ligh­ted the impor­tance of the rapid adop­tion of gene­ral-pur­pose tech­no­lo­gies for mili­ta­ry needs, and their effec­ti­ve­ness on the battlefield.

In dri­ving the reform of its defence inno­va­tion eco­sys­tem, the French Minis­try of Defence has, in recent years, taken on a more entre­pre­neu­rial and stra­te­gic role, with a more expe­ri­men­tal approach to inno­va­tion poli­cy¹. Never­the­less, the ques­tion remains as to the capa­ci­ty of our defence indus­try and our armed forces to be able to rapid­ly adopt all these gene­ral-pur­pose tech­no­lo­gies (GPTs) – such as arti­fi­cial intel­li­gence or quan­tum tech­no­lo­gies, at very dif­ferent levels of development.

Gene­ral-pur­pose tech­no­lo­gies can be defi­ned as those capable of conti­nuous future impro­ve­ments and with the poten­tial to serve as the basis for com­ple­men­ta­ry inno­va­tions in rela­ted appli­ca­tion areas (Teece, 2018). As such, a gene­ral-pur­pose tech­no­lo­gy has such an impact on ove­rall pro­duc­ti­vi­ty growth that it becomes ubi­qui­tous, as was the case with the inter­net in the 2000s.

With more and more prac­ti­cal appli­ca­tions in all sec­tors fol­lo­wing signi­fi­cant work in basic research, arti­fi­cial intel­li­gence (AI) has the poten­tial to become a gene­ral-pur­pose tech­no­lo­gy and thus trans­form the eco­no­my, natio­nal secu­ri­ty, and socie­ty as a whole.

Armed forces around the world are increa­sin­gly loo­king to AI as a key tech­no­lo­gy for their long-term stra­te­gies and plan­ning. With tech giants such as Google and Ama­zon at the fore­front of AI inno­va­tion, major US defence com­pa­nies are under pres­sure to step up their AI acti­vi­ties to keep pace with the com­mer­cial market.

100 mil­lion a year in arti­fi­cial intel­li­gence for defence, this repre­sents only 0.6% of the French armed forces’ equip­ment bud­get, whe­reas the Ame­ri­cans invest four times as much for an equi­va­lent bud­get².

Beyond invest­ment, seve­ral bar­riers to the adop­tion of these gene­ral-pur­pose tech­no­lo­gies per­sist. From a tech­ni­cal point of view, the vul­ne­ra­bi­li­ty of the sup­ply of real data to train the algo­rithms (Oso­ba and Wel­ser, 2017) is often the cause of unsa­tis­fac­to­ry per­for­mance. Howe­ver, the pro­blem of the confi­den­tia­li­ty of mili­ta­ry data has often, at least in dis­course, been an argu­ment that can hide ano­ther pro­blem : that of the gover­nance of data, which is too often left to the indus­trial owners of plat­forms and wea­pon systems.

AI can have a cri­ti­cal role in ope­ra­tio­nal plan­ning and decision-making.

In terms of user awa­re­ness, there are still other obs­tacles, often lin­ked to the une­qual know­ledge of deci­sion-makers of the uses of data and their added value for mili­ta­ry ope­ra­tions. This can be explai­ned by the fact that sys­tems inte­gra­ting AI do not all enter the tech­ni­cal reper­toire of the armed forces at the same time and with the same effi­cien­cy, but also by cultu­ral and orga­ni­sa­tio­nal bar­riers bet­ween the defence sec­tor and the civi­lian sec­tor. A strong disin­cen­tive to take risks, as well as pro­cu­re­ment pro­cesses that are not adap­ted to this type of tech­no­lo­gy, are often enough to dis­cou­rage any ini­tia­tive. Ulti­ma­te­ly, accor­ding to the figures pre­sen­ted by the Grand Défi focu­sed on arti­fi­cial intel­li­gence, only 10 to 15% of AI-based proofs of concept (in the sense of a demons­tra­tor) in France are indus­tria­li­sed and go to scale. 

AI can pro­cess large volumes of com­plex data to speed up deci­sion making at all levels of com­mand, acting as a force mul­ti­plier for com­mand – par­ti­cu­lar­ly with the emer­ging requi­re­ments of mul­ti-domain ope­ra­tions. While human resources can cur­rent­ly pro­cess, at best, 20% of the infor­ma­tion pro­du­ced today, this per­cen­tage may drop to as lit­tle as 2% in the face of the explo­sion in data pro­duc­tion³.

As a sign that this use case is of major inter­est to the armed forces, the Minis­try of the Armed Forces has just signed a defence and secu­ri­ty contract worth €240 mil­lion, over seven years, with a non-tra­di­tio­nal player in the defence indus­try : a pure player in defence AI in France. The aim of this contract is to acce­le­rate the intel­li­gence cycle and to pro­cess the tsu­na­mi of data coming from these sove­rei­gn CSO opti­cal ima­ging satellites. 

In addi­tion to hand­ling the expo­nen­tial growth of data, arti­fi­cial intel­li­gence can also improve the qua­li­ty of deci­sion-making pro­cesses and become a key player in the plan­ning and conduct of ope­ra­tions conti­nuum. Indeed, the aggre­ga­tion of many data sources, and their ana­ly­sis by machine lear­ning algo­rithms, can help deter­mine the best pos­sible geo­gra­phic dis­tri­bu­tion of forces based on the mis­sion, a unit’s capa­bi­li­ties, condi­tions in the area of inter­est, resup­ply requi­re­ments and infor­ma­tion deri­ved from the ana­ly­sis of all intel­li­gence sources.

Eve­ry theatre of ope­ra­tions offers an infi­nite num­ber of force confi­gu­ra­tions. To get an idea of the magni­tude of this num­ber, consi­der a game of chess. After each player has made four moves, there are 988 mil­lion dif­ferent pos­sible confi­gu­ra­tions. Com­bi­ning mas­sive sources of rele­vant infor­ma­tion (such as loca­tion data, wea­pon ranges, intel­li­gence, etc.) gives AI a cri­ti­cal role in ope­ra­tio­nal plan­ning and decision-making.

AI can go beyond human capa­bi­li­ties and dra­ma­ti­cal­ly reduce the res­ponse times of defence sys­tems to attacks by fast-acting wea­pons sys­tems (hyper­so­nic mis­siles, cyber-attacks or direc­ted ener­gy wea­pons) to bring about a radi­cal change in capa­bi­li­ties. Again, there are non-tra­di­tio­nal players in the defence indus­try, such as the US com­pa­ny Andu­ril, one of the few US start-ups to have suc­cess­ful­ly sca­led up. Its algo­rithms on board drones for tar­ge­ting pur­poses enable the US army to acce­le­rate the deci­sion loop at a tac­ti­cal level and to get clo­ser to real time to coun­ter ene­my manoeuvres much more effectively.

On so-cal­led broad-spec­trum effects, AI can also help to obtain a broa­der know­ledge and unders­tan­ding of the bat­tle­field, and for example anti­ci­pate attempts to mani­pu­late citi­zens. Indeed, AI-dri­ven ana­ly­sis of the effect on public sen­ti­ment of a kine­tic strike on infra­struc­ture, or AI-based solu­tions to detect dis­in­for­ma­tion cam­pai­gns, are among the most advan­ced intel­li­gence ser­vice tools for digi­tal influence and infor­ma­tio­nal warfare. 

Concrete appli­ca­tions of AI for the armed forces do exist, and not all use cases of this “enabling tech­no­lo­gy⁴” have yet emer­ged at this stage. Non-tra­di­tio­nal players in the defence sec­tor stand out as sup­pliers of these AI-based solu­tions, thanks to their abi­li­ty to deal with spe­ci­fic “user” needs and to pro­duce demons­tra­tors in “agile” mode based on expe­ri­ments. Howe­ver, these players are faced with the more gene­ral dilem­ma of adop­ting gene­ral-pur­pose tech­no­lo­gies : to what extent and at what speed do these players suc­ceed in chan­ging the tra­di­tio­nal forms of orga­ni­sa­tion of the armed forces to enable them to bene­fit from the advances pro­vi­ded by gene­ral-pur­pose technologies ?