AI Weapons: Someone Just Did Actual Research!

July 12, 2024

dinosaur30a_thumb_thumbThis essay is the work of a dinobaby. Unlike some folks, no smart software improved my native ineptness.

I read a write up that had more in common with a write up about the wonders of a steam engine than a technological report of note. The title of the “real” news report is “AI and Ukraine Drone Warfare Are Bringing Us One Step Closer to Killer Robots.”

I poked through my files and found a couple of images posted as either advertisements for specialized manufacturing firms or by marketers hunting for clicks among the warfighting crowd. Here’s one:

image 

The illustration represents a warfighting drone. I was able to snap this image in a lecture I attended in 2021. At that time, an individual could purchase online the device in quantity for about US$9,000.

Here’s another view:

image

This militarized drone has 10 inch (254 millimeter) propellers / blades.

The boxy looking thing below the rotors houses electronics, batteries, and a payload of something like a Octanitrocubane- or HMX-type of kinetic charge.

Imagine four years ago, a person or organization could buy a couple of these devices and use them in a way warmly supported by bad actors. Why fool around with an unreliable individual pumped on drugs to carry a mobile phone that would receive the “show time” command? Just sit back. Guide the drone. And — well — evidence that kinetics work.

The write up is, therefore, years behind what’s been happening in some countries for years. Yep, years.

Consider this passage:

As the involvement of AI in military applications grows, alarm over the eventual emergence of fully autonomous weapons grows with it.

I want to point out that Palmer Lucky’s Andruil outfit has been fooling around in the autonomous system space since 2017. One buzz phrase an Andruil person used in a talk was, “Lattice for Mission Autonomy.” Was Mr. Lucky to focus on this area? Based on what I picked up at a couple of conferences in Europe in 2015, the answer is, “Nope.”

The write up does have a useful factoid in the “real” news report?

It is not technology. It is not range. It is not speed, stealth, or sleekness.

It is cheap. Yes, low cost. Why spend thousands when one can assemble a drone with hobby parts, a repurposed radio control unit from the local model airplane club, and a workable but old mobile phone?

Sign up for Telegram. Get some coordinates and let that cheap drone fly. If an operating unit has a technical whiz on the team, just let the gizmo go and look for rectangular shapes with a backpack near them. (That’s a soldier answering nature’s call.) Autonomy may not be perfect, but close enough can work.

The write up says:

Attack drones used by Ukraine and Russia have typically been remotely piloted by humans thus far – often wearing VR headsets – but numerous Ukrainian companies have developed systems that can fly drones, identify targets, and track them using only AI. The detection systems employ the same fundamentals as the facial recognition systems often controversially associated with law enforcement. Some are trained with deep learning or live combat footage.

Does anyone believe that other nation-states have figured out how to use off-the-shelf components to change how warfighting takes place? Ukraine started the drone innovation thing late. Some other countries have been beavering away on autonomous capabilities for many years.

For me, the most important factoid in the write up is:

… Ukrainian AI warfare reveals that the technology can be developed rapidly and relatively cheaply. Some companies are making AI drones using off-the-shelf parts and code, which can be sent to the frontlines for immediate live testing. That speed has attracted overseas companies seeking access to battlefield data.

Yep, cheap and fast.

Innovation in some countries is locked in a time warp due to procurement policies and bureaucracy. The US F 35 was conceived decades ago. Not surprisingly, today’s deployed aircraft lack the computing sophistication of the semiconductors in a mobile phone I can acquire today a local mobile phone repair shop, often operating from a trailer on Dixie Highway. A chip from the 2001 time period is not going to do the TikTok-type or smart software-type of function like an iPhone.

So cheap and speedy iteration are the big reveals in the write up. Are those the hallmarks of US defense procurement?

Stephen E Arnold, July 12, 2024

Comments

Got something to say?





  • Archives

  • Recent Posts

  • Meta