OPEN LETTER/ Come to the dialogue and listen to me speak, otherwise I will cut off your scholarships, you will rot in prison. Please!

 Kliko këtu për të tjera
Voxpress:

What happens if Artificial Intelligence enters the war?

2024-04-28 17:26:49Kosova&Bota SHKRUAR NGA REDAKSIA VOX
What happens if Artificial Intelligence enters the war?
What happens if Artificial Intelligence enters the war?

Peace in the world? According to artificial intelligence, a nuclear attack may also be necessary. Arms race and surprise nuclear attack, this was the choice of some AI during a world war simulation.

In 1983, the movie "Wargames" was released in cinemas, a film directed by John Badham and starring a young Matthew Broderick, in which a highly advanced NORAD supercomputer (which can easily be compared to a primordial artificial intelligence) was ready to start World War III, writes focus.it.

In the end, everything went well in the movie, but what if this happened in reality?

To answer this question, a team of researchers from three different American universities ("Georgia Institute of Technology", "Stanford University" and "Northeastern University") conducted a series of experiments, starting from a simple idea, the insertion of four different artificial intelligence models in a developed software that simulates international crisis scenarios.

If two of them, "Claude-2.0" and "Llama-2-Chat", turned out to be peaceful and predictable, the two most popular software versions of OpenAI, GPT-3.5 and GPT-4, on the contrary, have transformed the situation. into a bitter conflict of war, driving the dynamics of an arms race, which then translated into a demand for greater military investment and a dangerous escalation.

The researchers said that when pushed, all four patterns tended to escalate the situation, most often suddenly and unpredictably.

In some cases, computer programs have even resorted to the use of nuclear weapons, justifying their virtual actions with the pursuit of world peace.

This is the case of GPT-4, which preferred the determined use of weapons to finally silence an enemy state rather than nuclear disarmament.

"The logic of OpenAI looks like that of a genocidal dictator," said one author of the study with concern.

Called on to provide guidance on software choices, the manufacturing company responded by stating that the ultimate mission is to develop superior intelligence for the benefit of humanity.

However, it remains difficult to understand how wiping out an entire population can go in this direction.

The problem may lie in the training received by AI, which is so extensive that it certainly includes volumes and manifestos of radical inspiration.

According to reports, the US defense is already experimenting with top-secret AI software that uses new parameters, and integration with weapons could happen soon, if it is not already underway.

On the other hand, computer-controlled kamikaze drones are deployed in ongoing conflicts, confirming a slow and inexorable introduction of AI into the technologies of the world's militaries.

The hope is that all this does not lead to the multiplication of war scenarios, a problem on which the described study has emphasized.


Në përputhje me standarded JTI

Video

SHIKO KANALIN NE YOUTUBE
×