Small Machine Warfare, Revisited

Infiltration and Off The Grid

In warfare, armies will be implementing robots with two capabilities. One capability is to infiltrate the enemy’s infrastructure. The other capability is to go off the grid for long periods, in order to be able to evade detection and be independent of GPS.

This is a follow-up to my post on Small Machine Warfare. It takes into account the development of Large Language Models, which I have argued makes it easier to train robots. And it takes into account what we have learned from Israel’s pager attack.

Infiltration

In the earlier post, I wrote,

As weapons get smaller, countries may be able to infiltrate weapons inside enemy countries. Then any country might be vulnerable to attacks from small machines.

We saw an example of this with Israel’s pager attack on Hezbollah. This suggests a model for infiltration: one side in a conflict creates a component that will be used by the other side. But unbeknownst to the other side, this component has military capabilities.

For example, suppose that the West could ship a component that Iran uses in its nuclear program and this component could spoil Iran’s enriched uranium. This would prevent Iran from developing a nuclear weapon at least as effectively as bombing its facilities.

As another example, consider all of the components in our communications systems that come from China. If China were able to insert elements that could be switched on in order to sabotage communications, that would be a very powerful weapon.

In an active war, such as that between Russia and Ukraine, tiny robots that could sneak behind enemy lines and wreak havoc might be very useful. This is yet another example of infiltration.

There is potential for infiltration in the ability to use Large Language Models to impersonate a human being. Imagine an online character that gains the trust of enemy personnel. Or imagine an online character that imitates an enemy general, so that enemy soldiers reveal secrets to it and obey its instructions.

Off the Grid

Another way to make military robots more useful would be to enable them to operate off the grid. Such a robot would require powerful batteries. It also would require internally-stored maps, so that it would not depend on GPS.

I picture an off-the-grid robot communicating only intermittently with its home operators. Most of the time, it would be silent, making it difficult to detect by electronic means.

It might have a specific target, such as an enemy headquarters or defense installation. It would be able to reach that target without GPS, so that it would not help the enemy to jam the GPS system. It would have the ability to disable the target by focusing on some key vulnerability.

The Trade-off

There is a trade-off between infiltration capability and off-the-grid capability. For infiltration, the ideal is to be small. Perhaps a robot the size of a coin could get inside the enemy’s building and collect intelligence or inflict damage.

But to operate off the grid, a robot would have to be large enough to carry a lot of battery power and computing power. I am guessing that it might have to be close to the size of a human being.

Communications Warfare

In addition to making robots easier to program, Large Language Models make it easier to interpret the enemy’s communications, provided that they are intercepted. This places a premium on securing our own communications and on being able to listen in on the enemy’s communications. Encryption becomes vulnerable, and decryption becomes potentially powerful.

I continue to believe that 20th-century weaponry, such as large ships, tanks, and aircraft, is becoming more vulnerable and less potent. I would be investing more in robots and in communication technology.