Ukraine War Day #283: Russian AI And The War Of The Future

Dear Readers:

Artificial Intelligence (AI) is one of those over-hyped and scary things, triggering fears of Terminator-like wars of the future, where flying computers, programmed with AI algorithms, hunt down and murder humans for sport.

From sci-fi meme to reality?

In “Metropolis” mad scientist Rotwang creates Robot-Maria

People who are not familiar with the foundations (and the mathematics) of Computer Science often get confused, and are easily manipulated, on the topic of AI. But the concept has been around for decades, even centuries, certainly long before modern computers were physically created. Science-fiction writers in the early part of the 20th century imagined intelligent machines, even flying robots. Their dream became a reality when modern computers were invented. However, all of current Computer Science, even the most sophisticated Machine Learning algorithms are still based on the mathematical theories of Alan Turing. Barring the invention of some other, completely different type of automation [hybrid computers containing human neurons?], even the Skynet Terminator is, at heart, a simple Turing Machine. Unlike human brains, which are infinite, Terminator can only perform a finite (albeit enormous) number of functions and processes; and he can only do what his core programming allows him to do.

On the other hand, the algorithms have gotten much more complicated, and the amount of data that one can store and refer to, has become almost infinite in size. The way a lot of so-called “Machine Learning” works: The automaton already knows what the end-result is supposed to look like, and then “learns”, by trying a billion different things, to reproduce that result. The analogy would be a human solving a jigsaw puzzle. You see on the box what the picture is supposed to look like, and then you keep trying to place all the pieces, until you have achieved the already-known result. If you solved the same puzzle over and over again, a million times, then you could end up skipping a lot of steps, as you learn. But it is still basically a mechanical process. And that’s what Machine Learning is.

A computer algorithm approximates the face of Alan Turing

The results of these computer innovations are awesome: For example, if you are attending a zoom-meeting at work, and your webcam is on, you can ask the zoom AI to draw a fake moustache on your face; and somehow it knows exactly where the contours of your nose are. I think even Alan Turing, were he to come back to life, would be impressed by this. Or, maybe he would just shrug and say, “Facial recognition, jolly good! They must have an infinite supply of punch cards or tape.” Because it’s still basically just a tape moving back and forth on a tape-reader. He who designed and punched the code on the tape, that was the human being. That was Rotwang. Or Alan Turing.

Before continuing on to today’s story from the Russian press, I just want to make one other small point: People often get confused between the words Automation and Artificial Intelligence. Automation is when computers do something that humans could already do by themselves, and in fact do it almost exactly the way humans do it, but just faster. For example, in theory you could add up a series of thousands of numbers, just using pencil and paper (or an abacus, or your fingers), but it might take you a long time. A computer, using counters and a loop, can do the same job in the wink of eye. But it still has to go through the steps. One simple distinction is this: When we talk Automation, we’re talking mostly digital (using discrete data); whereas AI deals with analog processes, continuous data, and non-discrete data. AI encompasses sub-branches such as facial recognition, voice recognition, pattern recognition, statistical control processes, and translation/interpretation of human languages. In the piece I am about to review, they are talking about the geometry and calculus of rocket trajectories. And of mathematical equations which a human being can actually perform manually, just not fast enough to get the job done. Therefore, I would qualify this as Automation rather than AI; however the interlocutor calls it AI. Admittedly, the line gets blurrier and blurrier.

Plug-Ins vs HIMARS

Reporter Evgeny Pozdnakov interviews a man named Yury Knutov, who heads the Air Defense Museum of the Russian Armed Forces. We have met Mr. Knutov before, which is why I have this very nice medallion of his face already in my Media library.

Yury Knutov

Knutov: “Initially, when the American HIMARS first appeared in Ukraine, we did not know how to fight against them. We decided to try the BUK-M3 complexes against them. This BUK complex is modern and completely digital, factually it is a computer with elements of Artificial Intelligence.

“Speaking about the HIMARS, it is important to note one of their characteristics: The rocket initially rises up to 20 km and only then dives down onto the target with a high degree of accuracy. The BUKs managed to intersect it exactly at a height of 20 km, they would then capture it on their radar screen, and then strike it. Currently we have collected even more data about these American devices. With the help of our data collection, we have been able to create a new [computer] program, which was added on to our other air-defense complexes. And now they have all been enhanced and automated [with this code plug-in].

“The problem earlier was that we did not have all this HIMARS data programmed into our Air-Defense [algorithms]. But now we have succeeded in learning, in detail, the parameters of the flight, and the points of intersection.

“Hence, this new knowledge that we acquired, is very important. Before, it was a manual process to capture the rocket on radar, and this very much complicated the work of our operators. But now we have eliminated the human factor in the process of fighting against the HIMARS. We are talking about unbelievable speeds, this is why, when working with these rockets, we need to use computers. Human reaction time isn’t quick enough to deal with it. We’re talking about milliseconds. Therefore, the appropriate algorithms have been added to many of our [Russian] air-defense complexes; and this has allowed us to greatly improve their effectiveness.

“In conclusion: The appearance, on the Ukrainian front, of these new weapons systems, including the GBU-39 bomb with the HIMARS engine, which is produced by Boeing, has provided us with much valuable data, while studying them. We will continue to work, to research, to master the statistical data. And, in the final analysis, we will perfect our algorithms, and these plug-ins will be added to all of our air-defense systems and complexes. In truth, this type of practice has been going on for decades, there is nothing new here.”