This is a synopsis of the cover article How Artificial Intelligence Can Revolutionise Science in The Economist on September 16.
The reason I want to introduce this article is first because I like the cover picture of this issue so much, and secondly, the acceleration of AI in the field of scientific research cannot be underestimated.
Here are the main contents.
Some people focus on the potential benefits of AI. By accelerating the pace of science and technology, especially in medicine, climate and environment, it may help humanity solve its most difficult problems. Will a golden age of science, technology and invention begin?
Inventions such as the telegraph, airplanes and the Internet once brought hope of world peace and the elimination of inequality, but it was just an illusion. History tells us that new methods and tools inspire scientific discoveries and innovations that change the world.
In the 17th century, microscopes and telescopes opened the eyes of researchers; scientific journals gave them a platform to share and publish. As a result, astronomy, physics and other fields made rapid progress. From the 19th century, the establishment of research laboratories led to the birth of artificial fertilizers, drugs and transistors (the basic elements of computers). In the 20th century, computer-based simulation and modeling made aircraft and accurate weather forecasts possible.
The computer revolution is not over yet. AI is now used in various fields of science, such as identifying factors with analytical value; finding patterns in large amounts of data; or modeling and analyzing complex systems such as protein folding and galaxies.
Two types of methods are promising for making new discoveries. One is literature-based discovery, which searches for neglected hypotheses, connections, and ideas in existing scientific literature. This can promote innovation across disciplines and at the margins of disciplines. The other is a machine scientist (self-driving laboratory) that analyzes existing data and literature, forms new hypotheses, and then tests them through a large number of experiments. Compared with human scientists, it is easier to replicate on a large scale and is less affected by bias.