A aviator doubles the checks of an assistance shipping to Ukraine in 2022 (photo of the Air Force by the staff sergeant Marco A. Gomez)
Washington – With Flux Foreign Policy, new army contracts in question and frozen humanitarian aid, including in Ukraine, there has never been pressure on the Pentagon to prove that military aid in Kyiv – Nearly $ 70 billion to date – is managed effectively and responsible.
The part of the lion of this burden is the command of army equipment which supervises everything, barrels of arsenals construction cannon, to equipment warehouses transferred under the presidential election of Drawdown, to an international network of distribution covering Private companies, NATO allies and front -line Ukrainian brigades.
American logisticians are rightly famous for their ability to move “iron mountains” of equipment from around the world. But their competence to keep a trace is uneven, as pointed out in the own inspector general of the Pentagon in several recent audits of help to Ukraine. So, make sure that the iron mountains arrive in good condition in the right place and on time requires the management of a mountain of documents. Nowadays, most of this data are virtual – but barely all.
“Inside the army, I would say that almost 100% of this data is automated,” said Chris Hill, data and analytical director of AMC. In other words, he told the defense of the defense in an interview, the information is automatically transmitted digitally from the database to the database without the need for a human to laboriously restore it to repainting it.
But everyone’s data? Not so much.
Kyiv’s post-Soviet institutions are still struggling to implement automation processes for modern commercial processes such as SAP, said Lieutenant-Colonel Colby Smithmeyer, MAPE analyst of the AMC. So, he told Breaking Defense: “The data we get from Ukrainians, this could come from a telephone call from a liaison officer, it could be an Excel for figures, or he could It is an printed Excel spreadsheet. “”
Standard data systems in the Department of Defense have never been built to tame such crazy men’s menagerie. Thus, the soldiers and civilians of the AMC innovated, improvised and, on occasion, have squarely between together a digital toolbox, including the analysis of big data, generative artificial intelligence and, in their Last experience, the blockchain.
“It is quite important that we maintain the visibility, responsibility and audit of this specific equipment,” said Pat Sullivan, a highly qualified civil expert in command, in an interview with Breaking Defense. “We must maintain this complete end (visibility) in a disputed environment, (therefore) we can see it and trace it and follow it throughout.”
Related: Télémane and Stocks: Army Materiel Command takes its own lessons from Ukraine
American warehouses at the Ukrainian fronts
Any given combat vehicle, ammunition box or supplies palette can start its trip in one of the AMC’s own arsenals, the warehouse of another government agency or the factory of an entrepreneur, explained managers during interviews. Then, the aid must be transmitted to military transport units and / or to the private maritime companies which move it physical -Star theater command. It was only then that he reached Ukrainians.
As a shipment crosses the step by step system, each of these actors must be able to see what happens to them. Then, once they have received it, they have to update everyone that they arrived intact and in their hands clean, more when and how they will send it. AMC therefore needs a system where all authorized actors – but no one else – can read and modify the data relevant to them – but not something else.
To follow all these transactions safely, AMC is testing a better known technology for cryptocurrency: blockchain. Like crypto, after all, information on military aid must be safely accessible by authorized users from around the world who do not share a single centralized database (and which may not be fully done trust).
“This proves traceability by creating an immutable transaction recording that allows the DOD to follow these assets,” said Major Matthew Goyette, who led a blockchain pilot exercise. “This removes all manual processes, Excel sheets, powerpoints, calls, emails. The data is chopped cryptographically in this “block” … so that some parts see it at the exact level you want them to see them. »»
The officials refused to say when the pilot started and how long he lasted, but they said that the AMC has exploited the expertise on the blockchain of a multitude of partners in and out of government: Defense Logistics Agency , Air Force Research Lab, Digital Manufacturing HUB MXD and Entrepreneur Simba.
But they could not simply copy existing commercial solutions, which consume a lot of treatment and electricity power, underlined Sullivan and Goyette. Instead, they built a slim blockchain below that only stored encrypted keys to logistics tracking data, rather than putting all the files in the blockchain itself.
What is the next step? “We have proven the initial pilot and determine with leadership how we play (UP),” said Sullivan. The future of AMC technology will depend on funding, he admitted. But it is optimistic because the order has invested resources and high -level attention in many digital projects, whose blockchain is only the last. Much of this investment has been motivated by the requests for the abolition of Ukraine.
Ukraine “lit a fire”
With its vast logistical responsibilities, the army of equipment has been strongly interested in better data management for years. But “once the Russians have invaded (Ukraine), it really lit a fire under this,” said Hill.
“We went there with algorithms that were not really” AI “. I would call them “Advanced Analytics,” he told Breaking Defense. “We decided that we have to bring more AI in this area, (and) about a year ago, we started to really enter the generative space.”
While chatbots are sadly famous for “hallucinating” invented information, an intensive study has reassured the Pentagon that the Genai can be forced to name only authority government data, instead of Reddit’s random publications on the collection of Pizza. Once properly tamed, large language models show a remarkable ability to digest and organize what is called “unstructured” data: information that is interspersed with an ordinary English text, rather than carefully categorized and labeled in a “structured” database.
This capacity is particularly practical for the AMC, because the orders of the presidential Drawdown authority specifying the aid to give to Ukraine are in fact not structured. Instead of sending precision formatted files which are transparent in a transparent manner in the military logistics databases, the White House emits PDF.
At the beginning, the army staff had to simply sit, read each order by line and hand all the relevant information in the appropriate fields in the AMC databases, said Lieutenant-Colonel Joseph Lavalle-Rivera . But the manual entry of data is notoriously at the high intensity of labor, slow and subject to error, and the withdrawal orders were simply too important, and the calendar too tight, to do everything by hand.
Even in the first orders, Lavalle-Rivera said: “You would have 200 articles with 12 to 15 different fields associated with each.“It is easily more than 2,500 individual entries in order of withdrawal – and they continued to grow.
The next AMC has tried to automate data entry by writing non -traditional non -AI code to search for drawing orders, extract data and download it. It worked better but only briefly, because the traditional IF-thel-Else software had trouble analyzing even minor formatting changes from one PDF to another.
“The format changes constantly, and each time they change the format, add a new piece of data, or anything... It immediately broke (the code), ”said Lavalle-Rivera to Breaking Defense. “Even changes to an additional tab, additional lines, can throw things like that.”
Thus, rather than rewrite and constantly rewrite the code to analyze PDF, AMC turned to Genai.
At first, it didn’t work well either. Uplining, the LLM did not “hallucinated” and did not add invented data because it was limited to the answer according to what was really in official documents. Downwards, it’s often omitting The data which was really there, because the algorithm initially interpreted the elements marked “classified” as things which he was not allowed to speak to its user.
Lavalle-rivera had to be lavalled “approximately a week of overdraft” before the LLM gave a full and precise response, he said. “This is largely limited to fast engineering. … There is more art than you think.
Predictive data for future fighting
Once the LLM has been tamed and its results are double verification, it has provided an effective way to transform the PDFs of the presidential Drawdown Authority not structured into structured data digestible by the Big-Data data system ‘Army Material Command.
This system is a variant of the army scale, but modified to allow easier entry of non-American data from partners like Ukraine. The command calls it Apas, the predictive AMC suite Analytics – “predictive” because it does not only follow the flow of weapons and supplies, but tries to predict what Ukraine (or American units) will then need.
“If you need ammunition and you are in the east of the Ukrain,” said Hill at Breaking Defense. “(Consequently) the importance of predictive analysis: I have to know when I have a problem a few months so that I can work this now.”
The American army learns a lot about the difficult logistics of modern war as it supports Ukraine, added Smithmeyer.
Start, he said: “We did not have a good idea for many of these consumption (rate) because we did not have a good reference base.” Take the rhythm to which the M777 Shutcuts wore their barrels of firearms, he proposed for example. “The shooting rate was different from everything we have certainly seen in Iraq and Afghanistan.”
Ukraine pours increasingly large quantities of combat data in the real world in the analysis of the army of equipment, said Smithmeyer.
“Now we follow the status of maintenance of the (Ukrainian brigades at the theater, we stick the consumption of supplies in the theater,” he told Breaking Defense. “(That) builds many sets of data that we can Now use… As we collect more and more data, we are able to obtain better and better predictions.
