Deepseek Technology is an ace of the Channel in China for the drone war

The Chinese release of the Deepseek chatbot sent Shock waves through the US stock market and beyond. Manufacturers claim that the underlying Deepseek-V3 model has been formed for around $ 6 million, a tiny fraction of the cost of rival American systems and with a tiny fraction of the computing power.

It happened just after the United States announced Massive expenses for AI infrastructure To build the type of computing power that Deepseek seems to make it questionable. The launch was described as “moment of Sputnik” Echo the events in 1957 when the Soviet Union shocked the world by taking an advance in the space race when it put the first satellite in orbit.

Spoutnik had obvious military implications. The Americans were nervously aware that Russian equipment was now flying above their heads every ninety minutes. A Chinese track in large -language models can look like a purely economic advantage. But a new generation of AI compatible drones could translate it into a military advantage.

Put an AI crown on drone supremacy

China is the world’s largest producer of small drones. Only one company, DJI of Shenzhen, commands an estimated 80% of the global drone market for consumer. Although the company officially deplore the military use of their productDJI drones have become omnipresent tools used by both parties in the war in Ukraine.

The low-cost FPV drones, armed racing quadcopters, may have become the most important unique weapon in the war on the ground. These are assembled locally, but many parts or most of the parts come from China.

At present, almost all drones are piloted by remote control. Qualified FPV pilots can mark remarkable successes: Tymofiy Orel, 25, was has awarded the title “Hero of Ukraine” To destroy 42 Russian tanks and 44 armored vehicles among other vehicles.

In the past year, small drones have been equipped with increasing AI levels. In particular, American manufacturers Auteri Supply Skynode-s advice in UkraineComputers the size of an optimized credit card for automatic learning applications, for example locking on a target and automatically controlling the best route. In the end, this could transform each drone operator into Tymofiy Orel.

In a Address in Rusi in OctoberFormer Ukrainian Defense Chief Valerii Zaluzhnyi noted how fast technology has evolved.

“When the robots entered the battlefield (in 2022), there were laughter of the Western press and local generals,” he said. “However, by 2024, technology, in particular artificial intelligence, began to play an important role in the war.”

The following year should see an increasing number of compatible drones AI on the battlefield in Ukraine. There are a certain number of challenges, especially in packaging enough calculation power on a small drone to run AI applications at high speed, and cause automatic learning systems to execute reliably and efficiently.

It’s just the kind of technology that China seems to have acquired with Deepseek.

From the virtual world to the battlefield

Deepseek is based on a large language model (LLMS), a specific type of AI formed on large quantities of text and excels in the realization of conversations in natural language. It may seem a world far from the challenge of driving a drone, but researchers have shown how LLM can improve drone operations. A large part of this research comes from China.

An article from 2024 offers a good overview of the domain – “Great language models for drones (non -mixed air vehicles): current state and future ways“, Was written by Shanghai -based researchers who describe how LLM integration with drones can improve autonomous data processing and rapid decision -making.

Although they carefully avoid specific applications, the authors note that “LLMs are exceptional to identify objects, individuals, vehicles or specific activities in video flows or images, providing detailed crucial information for military and civil surveillance operations. “”

This results in an ability to make complex orders such as “see if there are vehicles on the track near the river” or “” fly at low altitude in the enemy position and approach of the South. “”

The LLM allows the drone to accept orders, then report what he sees in natural language: “No vehicles in motion on the track, a truck burned as previously indicated” or “There was no fire Enemy positions, but heat sources show they are occupied.

In addition to communicating with humans, LLMs allow drones to talk to each other in natural language. A group of drones can share the results and coordinate its actions according to its mission and to distribute tasks without human intervention. A recognition drone can assign attack drones to specific targets, then send follow -up to those who are not destroyed.

Another Chinese newspaper of 2024 goes further – Inhabited air vehicles and without pilot frame-combat cooperative combat based on important language models Described the work using LLMS to make “completely autonomous driving planes”. A reality. They envisage a single human fighter pilot with a collection of robotic asients who react at the speed of the machine and communicate in human language. Again, they take advantage of the LLM planning and problem -solving capacity to make it an effective pilot, capable of responding quickly to unexpected situations. (It would be weird but not surprising if the drones imitated human practice and called themselves as “Maverick” and “Iceman”).

It is interesting given this capacity that Deepseek-R1 is open-source. The Chinese are probably very aware of its double -use potential. Are there any measures in place to prevent it from being used by an opponent? Or do they have something even more powerful reserved for their own use?

Take advantage of the in -depth advantage

Deepseek is supposed to be have specific advantages On other LLMs, in particular that it is economical in resources, which means that it needs less computing power than other models and adapted to EDGE applications as drones rather than for data centers . In addition, the developers claim to have obtained results quickly with less training data than comparable models.

We don’t know how much Deepseek is really capable. But a powerful and low cost system that can be deployed on drones would be a useful asset for China at this stage.

China has the largest drone manufacturing basis in the world. They have a multitude of information on the use of these drones in combat with tactics and techniques developed by Ukraine and Russia in a war that has cost them nothing. And they were recently Develop their own drone combat units Equipped with what looks like low -cost commercial drones.

Add an LLM capacity to the mixture and China seems to contain all the cards to build an unbeatable drone force. War is becoming more and more a question of Drone VS Drone. In the long term, the software can prove to be the decisive factor because the victory will go aside with the smartest drones.