Until now, it seems that the world could live more or less in peace. There was life insurance, called New START which allowed experts from the United States and Russia to know even the smallest details of the missiles that the other had.
But on February 5, all this expired. The treaty was agreed for a period of ten years, until February 5, 2021, extendable once for another five.
Now, for the first time in fifty years, the two largest nuclear powers on the planet have no rules limiting their arsenalsand, taking into account the world situation, the situation does not look too good.
The problem is that, although they say that for the moment they are going to maintain the status quoThe reality is that no one trusts anyone. Russia and the US are improving their arsenal; China is also joining this race and countries like South Korea are beginning to value the idea of having their own bomb.
An algorithmic police to control the world
In this scenario, a proposal has emerged. Matt Korda and his team at the Federation of American Scientists (FAS) have launched this idea: if we can’t send human inspectors, let’s send algorithms.
The plan is called Cooperative Technical Media and consists of use the network of satellites that already surrounds the Earth to monitor arsenals from space, letting artificial intelligence do all the analysis.
It is, as Korda himself says, in an article by The Wiredhe Plan B. It is not the best idea, but Plan A is there, to do it with humans through diplomatic means, no longer exists. The idea is that if you have a well-trained AI, you can use it to detect changes in a military base.
A necessary ‘Plan B’ if humanity wants to preserve itself as we know it
To give you an idea of how necessary all this is, in 1985, the world had more than 60,000 nuclear weapons; Today it is just over 12,000.
That drop of 50,000 bombs was not by magic, but by decades of diplomats and scientists reviewing each change. With the death of New STARTall that work could disappear and reach numbers higher than those of 40 years ago.
What the FAS proposes is to find a middle point. «If we cannot enter their bases, let them open the doors for us from above»says Korda.
The trick would be technological cooperation: one country could warn the other by saying that, to verify the treaty, on such a day at such a time it will open the hatch of a specific silo just as the satellite passes overhead.
However, Sara Al-Sayed of the Union of Concerned Scientists warns that it is not as easy as it seems. An AI needs data, and there is no public archive with high-resolution photos of all the missiles in the world. Specific databases would have to be created for each country.
Besides, All this is not just about missiles and it would also be necessary to track submarines, bombers, production plants and even test sites..
One of the biggest fears of experts is the ‘randomness’ and continuous failures of AI
You well know that ChatGPT, for example, sometimes hallucinates or that facial recognition systems fail. In arms control, a false positive could trigger a catastrophic response. A slight mistake can be the beginning of the end.
Despite all this, the proposal still stands because the alternative is nothing. It is not about AI leading to total disarmament, but about preventing the situation from getting out of hand. As Korda says, it is a triage measure.
So to speak, they prefer an imperfect solution and an algorithm that can potentially fail (humans will be there to control it) than to be completely in the dark while certain countries do and undo as they please with something so delicate.