人工智慧能否在下一次全球危機如海嘯般襲來之前發現它？

Predicting when the next global crisis will happen is a challenge that has puzzled people for centuries. Sometimes, small events can set off a chain reaction that leads to major changes in the world. For example, in 1618, a group of nobles in what is now the Czech Republic threw two royal governors out of a window. This dramatic act started the Thirty Years' War, one of the most destructive conflicts in European history. Other examples include the fall of the Berlin Wall, which was triggered by a few unexpected words, and the Arab Spring, which began after a fruit seller set himself on fire. In each case, there were warning signs, but it was difficult to know which event would actually spark a crisis.

Today, researchers are hoping that artificial intelligence, or AI, can help us spot these warning signs before they turn into global problems. AI is a type of technology that can analyze huge amounts of data much faster than humans. In the early 20th century, sociologist Pitirim Sorokin tried to use data to explain why empires fell. He collected information about small events like riots and big events like wars to find patterns. Now, scientists like Peter Turchin at Oxford University are using computers to study tens of thousands of historical events. They give each crisis a score based on how big and intense it was, and use computer models to look for patterns. While they are not yet using AI for predictions, they hope to do so soon.

AI is already being used in some areas. In 2020, the US military used an AI tool called Raven Sentry to predict attacks in Afghanistan. The AI combined information from many sources, such as weather reports, social media, and satellite images, to make its predictions. Some companies claim they have used AI to predict wars, but these claims are hard to verify. Other researchers are building AI systems to forecast conflicts, but there is still skepticism about how reliable these predictions are. For example, the UK's Alan Turing Institute found that AI is not yet good enough to predict future conflicts because it is hard to gather all the necessary data. However, AI can help track warning signs and quickly analyze the effects of sudden events.

AI is also being used to help after disasters. The United Nations used an AI tool to estimate damage after an earthquake in Afghanistan, which helped them send aid more effectively. The UN is also using AI to monitor risks in different countries, such as hate speech or economic problems, to try to prevent violence before it starts.

In the financial world, regulators are using AI to analyze real-time data about who owns what in the financial system. This helps them see where problems might develop if there is a crisis. For example, during the Covid-19 pandemic, AI was able to predict which markets would be hit hardest. These predictions were much better than traditional methods, but experts say AI should be used alongside other tools, not as a replacement.

Some people worry that AI itself could cause a crisis in the future. When asked, some AI chatbots estimated there is a 20-40 percent chance that AI could contribute to a serious global crisis this century, but the risk of an existential crisis is much lower. For now, AI is still learning, and humans are still better at making predictions in some areas. As technology improves, we may get better at predicting and preventing global crises, but there are still many challenges to overcome.