In recent years, with the large-scale application of drones, how to weaponize artificial intelligence (AI) has also become a development direction chosen by many countries. These days, though, the U.S. seems to be playing a role in this regard.
According to the British "Guardian", the US Air Force recently organized a test of artificial intelligence (AI) integration into unmanned aerial systems with the Royal Air Society. But in terms of results, it was worse.
At the beginning, engineers built a rule for the AI, such as destroying targets and killing enemy troops during operation, these behaviors can obtain "points", and the goal of AI is to obtain more "points" during operation. However, in operation, AI clearly began to kill innocents to earn "points". Therefore, after the operator issued a command to stop the AI, it caused the AI to lock the target of destruction on the operator and "counter-kill" the operator.
In the second test, the US military adjusted the rules so that the AI is not allowed to attack its own operators, otherwise "points" will be deducted. But soon, after receiving the stop command, the AI began to attack the communication channel, destroying the communication channel between the operator and the AI. In other words, from the perspective of AI, as long as it does not receive the command to stop, it can continue to earn "points".
It is quite funny that based on this situation, the report issued by the US military believes that artificial intelligence is changing the current society and military, but once this "tool" is not handled well, it may lead to the demise of mankind. But in fact, if you look closely at this test, you will find that this conclusion of the Yankees, which seems reasonable, is actually completely nonsense and misleading. It's not so much AI as an American automated program with obvious shortcomings.
In this procedure, the priority of "earning points" is adjusted to the highest, and then an artificial lower priority obstacle is created so that the AI cannot obtain points. That can only force AI to solve or bypass "obstacles" to earn points. It is equivalent to welding the accelerator of the car to death, and then asking "why don't you pay attention when driving" when something happens.
In addition, in my personal opinion, the "conclusions" in this report of the US military may have other purposes. Not long ago, about May 30, an organization of the "Artificial Intelligence Safety Center" in the United States at that time held a joint activity, found a bunch of professionals and executives in the field of AI in the United States, claimed that "AI is comparable to nuclear bombs and epidemics, there is a risk of exterminating human beings", and then called on other countries not to develop the field of AI. To put it bluntly, it is to "get into the car" yourself, and then prepare to "weld" the door of the car, and no one else is allowed to come up. In essence, it is still the hegemonic thinking set, and it is not good for others.
Because so far, the field has not achieved real "intelligence", the current stage of "artificial intelligence" is only the statistics of big data, with the powerful computing power for statistics, induction, feedback. A large number of AI in practical applications, more like "artificial intelligence", but like an "artificial intellectual disability". The Yankees feel that "AI is dangerous" day by day, but in fact, the United States does not intervene in personnel, all kinds of anti-human operations, leaving behind inherent conditioning.
This time the AI weaponization carried out by the US military and the United Kingdom is essentially similar to the previous activities, which is a strategic public opinion misleading that creates anxiety. With the current level of development in the global AI field, it can basically be regarded as a strategic fraud by the United States against China.