Listen at 38:00. Neural networks feast on Big Data. Military intelligence reports are the next target for machine intelligence. This is our future: the data silos of military intelligence will be united by AI. Neural networks will draw upon intelligence reports and their corresponding data, to build autonomous annotators. Satellite data and news reports, spies’ work, and military logistics data will all feed in, and relevant information will be flagged and described to us. Why does this matter? Because machine intelligence is far more efficient than our brains. 10,000 times more efficient.
Neural networks recognize images using a tiny fraction of the computation of our brains. As a result, we can already run brain-scale intelligence on a single existing supercomputer. China has the scale of computation necessary for human-level intelligence, and it is already running an entire city with AI. Heck, Google has enough spare processor power to run a super-human intelligence. The military does, too.
After a multi-modal neural network captures military data and runs on government hardware, the next step is simulation: feed the network false data, imagined scenarios, to see how it would react. Like DeepMind’s AlphaGo, the neural network would model military conflicts. It can’t pull the trigger, yet it generates the plausible scenarios that generals choose. AI won’t ‘run’ the military — it will simply give them the best suggestions. If AI is right more often than people, why not listen to it? It will become our strategic advisor.
With machines trained on military data, countries with superior compute capacity will be able to out-think and out-position opposing forces. Military activity will be more likely to succeed, and resolve faster. It will cost less, and gain more. Military activity will increase.
We have seen this pattern before: when navies acquired powerful cannons. It was called ‘Gunboat Diplomacy’. Superior weaponry and tactics made the use of force or the threat of force an easier route to political goals. If military success is assured, why not use the military to succeed! AI will provide a similar overwhelming advantage.
So, when this advantage arises, increased influence will go to the countries with the political will to utilize this strategy. Imagine if Putin invaded Ukraine with slaughterbots. China would gladly exert more military pressure on the South China Sea and the strait of Malacca. The US could reduce casualties from combat, which are our primary source of reticence during our military engagements.
Just as superior cannons led to continental conflicts and international military activity, AI will foster war by increasing the success and decreasing the cost to a few countries. Machines in charge of strategy will allow militaries to resolve conflict or disable rivals faster than ever. And, there will be great pressure to utilize that advantage, before others catch on. Ecclesiastes may be wrong — success is a race to the swift.
The surest advantage will go to the country which aligns its ‘strategy-bot’ with an ‘economy-bot’. Let the stock market and consumer market go as they please. Yet, military necessities can be imposed by the government when needed. The AI suggests when, how, and by how much. The generals agree, because the machine is usually right. The people at large agree, because they see success. I am describing China, not the US.
A planned economy is sluggish and slow to adapt, yet, an economy with the option to be planned can pursue strategies which are unavailable to either a market-driven economy or a pure planned economy. China’s ‘lever’ on government involvement gives it a unique strategic advantage. It isn’t a Stalin-esque pure planned economy — China is ‘planned on-demand’.
AI will be tasked with the design of military hardware, too. What drone chassis are best spec for each target and application? Machine intelligence knows the answer — it is better than us at designing reinforced structures and meeting multitudinous constraints. I suspect that neural networks will not prefer current components; tanks are obsolete, as are helicopters, because they are large targets. Aircraft carriers are not as threatening as a swarm of mechanical dolphins. Machine intelligence will prefer numerous, modest forces, which are more difficult to deflect or disable completely.
The surest way to conquer is to disable your rival’s infrastructure. Like General Sherman’s march, drones will be trained to target equipment and facilities, not people. This will provide the military with the surest cover against moralizations: “Our robots don’t ever shoot at people… just power lines and radio towers.” It’s okay for machines to pull the trigger, if they only attack other machines.
Robots also make ‘soft’ weapons more desirable. If a team of humans deploys a sound-cannon, or some other non-lethal weapon, it only subdues and distracts opposing forces. The sound-cannon operators risk their lives, if they don’t have lethal forces supporting them. In contrast, a robot with a sound-cannon is disposable. So, it is easy to deploy many non-lethal robots.
Robots with non-lethal weaponry allow for general suppression of a population — the robots can set-up and maintain a larger perimeter, as well as rapidly segment regions to cut enemy supply lines and movement of forces. Non-lethal robots quarantine quadrants, and ensure safe passage for your own forces.
So… if your power goes out, cell towers go down, and a perimeter of high-pitched sound-cannon quadcopters monitors everyone’s movements, then humans with tele-operation can pull the trigger of a gun with wheels. Soldiers are parked in an APC miles away, while their RC terminators rove the streets. Get ready, Delhi.