Chinese scientists, diplomats and intelligence experts packed together a nice dataset out of cocktail-party gossip, diplomatic cables, intelligence collected by various means, directives, OSINT, etc. according to a source (and other sources citing this sole source). And once they had the dataset, they let their previously developed AI to train on this dataset and it is now readily able to suggest directions for the human decision makers of the Chinese diplomacy (or intelligence services).
The article, which is obviously based on official info describes the phenomenon as “a new type of robot with the potential to change the game of international politics forever”. To sense what is at stake one doesn’t have to read between the lines, because it is all there: its task is to “devise a strategy that can be never thought of before by humans”.
So, already AI is used to circumvent and / or exploit human weaknesses – at an official level.
According to the article, there are a number of AI-driven decision support systems already in use by the foreign policy system of China. The main benefit of the AI -in the words of a scientist is that
“artificial intelligence systems can use scientific and technological power to read and analyse data in a way that humans can’t match”
“human beings can never get rid of the interference of hormones or glucose”
and also that
“the AI policymaker, however, would be immune to passion, honour, fear or other subjective factors. It would not even consider the moral factors that conflict with strategic goals”.
So this is how our ‘moral factors’, like passion, honour and fear suddenly become human weaknesses – to which a remedy must be found by the machines. And you know what? It’s just one step ahead to search for exploits, that are targeting these human incapacities.
For what we know, deploying AI into high-level foreign, military and intelligence affairs-related analysis and decision making is not quite new, it is being done all around the world since the 1960s with varying success.
What is news indeed is that it is the first time when an official outlet thinks that the AI could one day outpace its masters.
Diplomacy is a game of imperfect information, which means that the players are usually unaware of the actions chosen by other players – but they know who the other players are, what their possible strategies and tactics are and what the preferences/payoffs of these other players. But while China is having a centralized political scene, other societies are electing their leaders and this introduces randomness in the type of player with whom you play a game. Which means that the AI could never beat a human player, because could not play a game of imperfect information (for which it was designed), but a game of incomplete information – because the AI may not know some information about the other players, e.g. their “type”, their strategies, payoffs or their preferences. In other words: any dataset would lack some crucial elements of information.
And while the human beings demonstrated their abilities to solve puzzles without knowing all or sometimes enough, the AI so far demonstrated only that it can not function if the dataset they were trained on is different that of the real world. And it would be chilling to have an AI controlling the question of peace of war…
Not to mention how safe it is to feed confidential intel into a system cannot be understood – because of the two vulnerabilites of the AI: the interpretability problem (a.k.a. the “black box” problem) and the reproducibility problem of all artifical intelligence methods.
I for one am absolutely convinced that the researchers are already seen the elephant in the room, because they realized that nations one day could be run other than a human power elite.
So, for us the only thing that is important now is that it is reasonably assumable that the machine learning on past and present foreign affairs datasets has been completed in a number of countries and some related AI capabilities has been mated with a leader-useable interface.