The Military Plans to Let Swarms of Armed Drones Talk to Each Other

 Globally, there are now over 57,000 companies trying to sell you AI-driven products. It's the bandwagon everyone's jumping on. And for good reason. AI in some form will clearly change everything.

Currently, all these AI tools deliver astounding amounts of information at high speed. But they can't quite process data like a human brain. Our minds adapt instantly as situations change. We can be working on one thing and, if a new crisis hits, change focus with ease.

AI is limited in how it "thinks" by the computer chips it uses. Traditional CPUs and GPUs are fast but they're hard-wired -- they have fixed pathways that never change. Computer chips have none of the placidity of human neurons. Sure, AI can simulate brain-like behavior, but only by forcing enormous amounts of data down these fixed pathways and processing the results in ways that seem human. It's imitating our brain through brute force.

But there is one invention that might change that.

It's called a neuromorphic chip, and it's designed to work more like actual brains. Its architecture integrates memory and computation together, mimicking the all-in-one power of human neurons. And it leverages event-driven processing, firing only when an input occurs, just as neurons fire in response to stimuli. 

The results are powerful. A self-driving car no longer needs to process everything it sees in some enormous, never-ending stream of data. The neuromorphic chip allows the AI to react only when an event signals a change. Someone suddenly steps into the street? The neuromorphic pathway lights up, instantly making a decision (hopefully to not run that person over). And where the current AI processes one thing at a time, neuromorphic chips act in parallel, like human neurons, processing multiple problems at once. They are faster and more efficient.

The latest example of the neuromorphic chip in action is Intel's Hala Point system. Unveiled in April of 2024, it's the world's largest and most advanced neuromorphic computer -- and a huge leap in brain-inspired AI hardware.

The self-driving car might use one neuromorphic chip to process a quick decision. Intel's Hala Point uses 1,000 neuromorphic chips to simulate 1.15 billion artificial neurons, getting disturbingly close to the complexity of a primate's cortex. The system can learn in real-time, making complex problem-solving possible. Scientific advances, city infrastructure, global logistics -- the Hala Point can handle all kinds of new challenges.  It's like a human brain, but with 2,300 embedded x86 processors. It's a super computer that also happens to think like us. 

Intel took the neuromorphic model and went big. But researchers at Texas A&M University are breaking new ground by thinking small. Extremely small.

They plan to develop a neuromorphic chip the size of a grain of rice.

A tiny AI chip opens up all kinds of possibilities. Suddenly the smallest devices can leverage the power of AI. Because the chip contains both the memory and computation power, it doesn't demand the additional battery power of traditional chips, and small devices have no room for more batteries.

Imagine hearing aids that enhance speech in real time, wearable monitors that warn of heart attacks, or autonomous micro drones for disaster search and rescue. There are endless ways this could help humanity.

Of course, no one approached the scientists at Texas A&M with money to do any of those things. Instead, they were given a huge grant via the 2025 Air Force Office of Scientific Research Young Investigator Award. The revolutionary neuromorphic chips are targeted for use in military drones.

According to Professor Suin Yi, one of the project leaders at the University, "small drones don't have an engine, so their energy budget is very small. That's why it's very important to come up with a game changer such as neuromorphic computing systems beyond digital computers that allow battery-powered drones to keep the same fly-time with AI as those without."

Any one of these new drones would be no match for the MQ-9 Reapers currently being used by the Air Force, with their 66 foot wingspan, hellfire missiles and laser guided bombs.

But the Air Force isn't out to make one of them. They hope to outfit thousands of them. And the ultimate plan is for the micro drones to act together.

See, they've been working on technology to network groups of weapons. The project is code-named Golden Horde. It's an initiative focused on advancing collaborative and autonomous weapons capabilities. The goal is for autonomous bombs to share data, assess the battlefield, and coordinate attacks without human intervention.

It's a frightening idea, weapons talking to each other without humans in the loop. Up to now, Golden Horde has focused on the CSDB, or Collaborative Small Diameter Bomb. It's a kind of guided missile. The Air Force has never been able to bring this AI technology to a swarm of micro drones.

They were missing the right chip.