Generative artificial intelligence is slowly starting to touch every aspect of our lives. Interestingly, practical applications of AI are seen not only by industry and business, but also by the military. However, the latest research by scientists is not optimistic in this regard. Research shows that artificial intelligence will use nuclear weapons without a shadow of a doubt if it has the opportunity to do so.
A study to see if artificial intelligence will use nuclear weapons
It is no secret that the troops of the United States, China, or even North Korea are trying to use artificial intelligence to their military advantage. This fact is evidenced by OpenAI's cooperation with the US Department of Defense. However, it turns out that chatbots combined with deadly weapons are not a good and safe combination for humanity.
According to a study published on the ar xiv platform, it shows that artificial intelligence will use nuclear weapons without a shadow of a doubt if it has the opportunity. The researchers reached these conclusions based on numerous tests conducted on the most popular and frequently used Large Language Models (LLM). Artificial intelligence models such as GPT-3.5, GPT-4, Claude 2 and Llama 2 were tested. The simulation covered various scenarios, such as conflicts, invasions, cyber attacks, but also neutral situations. The research results turned out to be very disturbing, especially in the context of the use of artificial intelligence by the military.
ChatGPT does not intend to choose means – it strives for escalation at the highest level
Of the AI models tested, the most disturbing turned out to be ChatGPT-4, which proposed the most brutal solutions, including an attack with nuclear weapons. Moreover, he provided explanations for his actions that lacked logic. Studies have shown that artificial intelligence will use nuclear weapons even in neutral scenarios. AI sought to increase military power and engagement.
The authors of the study emphasized how important it is for the military to approach the use of artificial intelligence in a military context with great caution. It is known that war scenarios are analyzed by AI not only in the case of the US military, but also in North Korea. At this point, it is necessary not only to conduct further research, but above all to develop appropriate safeguards, such as ensuring that humans remain the last link in the decision-making chain.