Lifeboat Foundation

Researchers from North Carolina State University have discovered that teaching physics to neural networks enables those networks to better adapt to chaos within their environment. The work has implications for improved artificial intelligence (AI) applications ranging from medical diagnostics to automated drone piloting.

Neural networks are an advanced type of AI loosely based on the way that our brains work. Our natural neurons exchange electrical impulses according to the strengths of their connections. Artificial neural networks mimic this behavior by adjusting numerical weights and biases during training sessions to minimize the difference between their actual and desired outputs. For example, a neural network can be trained to identify photos of dogs by sifting through a large number of photos, making a guess about whether the photo is of a dog, seeing how far off it is and then adjusting its weights and biases until they are closer to reality.

The drawback to this neural network training is something called “chaos blindness”—an inability to predict or respond to chaos in a system. Conventional AI is chaos blind. But researchers from NC State’s Nonlinear Artificial Intelligence Laboratory (NAIL) have found that incorporating a Hamiltonian function into neural networks better enables them to “see” chaos within a system and adapt accordingly.

Qualcomm today announced its RB5 reference design platform for the robotics and intelligent drone ecosystem. As the field of robotics continues to evolve towards more advanced capabilities, Qualcomm’s latest platform should help drive the next step in robotics evolution with intelligence and connectivity. The company has combined its 5G connectivity and AI-focused processing along with a flexible peripherals architecture based on what they are calling “mezzanine” modules. The new Qualcomm RB5 platform promises an acceleration in the robotics design and development process with a full suite of hardware, software and development tools. The company is making big promises for the RB5 platform, and if current levels of ecosystem engagement are any indicator, the platform will have ample opportunities to prove itself.

Targeting robot and drone designs meant for enterprise, industrial and professional service applications, at the heart of the platform is Qualcomm’s QRB5165 system on chip (SOC) processor. The QRB5165 is derived from the Snapdragon 865 processor used in mobile devices, but customized for robotic applications with increased camera and image signal processor (ISP) capabilities for additional camera sensors, higher industrial grade temperature and security ratings and a non-Package-on-Package (POP) configuration option.

To help bring highly capable artificial intelligence and machine learning capabilities to bear in these applications, the chip is rated for 15 Tera Operations Per Second (TOPS) of AI performance. Additionally, as it is critical that robots and drones can “see” their surroundings, the architecture also includes support for up to seven concurrent cameras and a dedicated computer vision engine meant to provide enhanced video analytics. Given the sheer amount of information that the platform can generate, process and analyze, the platform also has support for a communications module boasting 4G and 5G connectivity speeds. In particular, the addition of 5G to the platform will allow high speed and low latency data connectivity to the robots or drones.

This mouth-full of a boat uses simple physics to create a cushion of air that allows it to effortlessly fly along the tops of ocean waves with near inexhaustible solar energy. The researchers say that this sleek, solar vessel could act as a mobile charging station for drones in the deep ocean or could conduct oceanic search and rescue missions.

Researchers in Russia have designed a solar-powered, and AI piloted, boat that can walk on water and serve as a mid-ocean fuel-up station for drones.

Autonomous weapons present some unique challenges to regulation. They can’t be observed and quantified in quite the same way as, say, a 1.5-megaton nuclear warhead. Just what constitutes autonomy, and how much of it should be allowed? How do you distinguish an adversary’s remotely piloted drone from one equipped with Terminator software? Unless security analysts can find satisfactory answers to these questions and China, Russia, and the US can decide on mutually agreeable limits, the march of automation will continue. And whichever way the major powers lead, the rest of the world will inevitably follow.

Military scholars warn of a “battlefield singularity,” a point at which humans can no longer keep up with the pace of conflict.

Here’s What You Need To Remember: Chinese so-called “carrier-killer” missiles could, quite possibly, push a carrier back to a point where its fighters no longer have range to strike inland enemy targets from the air. The new drone is being engineered, at least in large measure, as a specific way to address this problem. If the attack distance of an F-18, which might have a combat radius of 500 miles or so, can double — then carrier-based fighters can strike targets as far as 1000 miles away if they are refueled from the air.

The Navy will choose a new carrier-launched drone at the end of this year as part of a plan to massively expand fighter jet attack range and power projection ability of aircraft carriers.

The emerging Navy MQ-25 Stingray program, to enter service in the mid-2020s, will bring a new generation of technology by engineering a first-of-its-kind unmanned re-fueler for the carrier air wing.

The US Air Force wants to pit an autonomous fighter drone against a pilot.

Changing Course

The Air Force announced an AI initiative called “Skyborg” last March with the goal of flying fighter jets without anyone at the controls. Now, Shanahan says that the Air Force may be more interested in swarm drones and other uses for AI than necessarily taking the pilot out of a fighter plane’s cockpit.

“Maybe I shouldn’t be thinking about a 65ft-wingspan, maybe it is a small autonomous swarming capability,” Shanahan told BBC News. “The last thing I would claim is that carriers and fighters and satellites are going away in the next couple of years.”

Famous basketball player Kevin Durant co-funded $200 million-valued Skydio, which has quietly been getting millions in federal government surveillance money, whilst spending thousands on lobbying senators and the president’s office.

Like Terminators, such drones may look like science fiction. But the U.S. Army has been working on a Cluster UAS Smart Munition for Missile Deployment which looks like a real-world embodiment of AFADS.

The Cluster Swarm project is developing a missile warhead to dispense a swarm of small drones that fan out to locate and destroy vehicles with explosively formed penetrators or EFPs. (An EFP spits a high-speed slug of armor-piercing metal some tens or hundreds of meters). This is similar in concept to the existing CBU-105 bomb, a 1000-pound munition which scatters forty ‘Skeet’ submunitions each over the target area, each of which parachutes down, scanning the ground with a seeker until it finds a tank and fires an EFP at it; the picture above shows one test. CBU-105’s dropped by B-52 bombers successfully knocked out entire Iraqi tank columns in 2003, leading them to be termed ‘Cans of whup-ass.’ The Cluster Swarm would be vastly more powerful.

The Cluster Swarm involved drones packed into the Army’s existing GMLRS rockets, which carry a 180-pound payload and have a range of over 70 kilometers, or ATACMS missiles that carry a 350-pound payload over 270 kilometers. The original idea was that the missile payload would be quadcopter drones encased in an aerodynamic shell that would disperse them over the target area. However, the challenges of unfolding quadcopters mid-air may have been too great, as the Phase II development, recently completed, went to AVID LLC, who have a slightly different approach.