There is an old A joke that physicists like to tell: everything has been discovered and reported in Russian journals in the 1960s, but we don’t know it. Although exaggerated, this joke accurately captures the current state of affairs.Huge amount of knowledge and rapid growth: The number of scientific articles expected to be published on arXiv (the largest and most popular preprint server) in 2021 Reached 190,000——This is only a subset of the scientific literature produced this year.
Obviously, we don’t really understand what we know, because even in their own narrow field, no one can read all the literature (except journal articles, doctoral dissertations, laboratory notes, slides, white papers, technical notes, and report). In fact, in this mountain of papers, the answers to many questions are hidden, important discoveries are ignored or forgotten, and connections are still hidden. This is entirely possible.
Artificial intelligence is a potential solution.Algorithms can already analyze text without human supervision to find help to reveal the relationship between words knowledgeHowever, if we stop writing traditional scientific articles that have hardly changed in style and structure in the past 100 years, then we can achieve more.
Text mining has many restrictions, including access to the full text of the paper and Legal Issues. But most importantly, artificial intelligence has not really Understand the concept And the relationship between them, and is sensitive to biases in the data set, such as the choice of papers it analyzes. Artificial intelligence—in fact, even non-professional human readers—have difficulty understanding scientific papers, partly because the use of jargon varies from discipline to discipline, and the same term may have completely different meanings in different fields. The increasing interdisciplinary nature of research means that it is often difficult to use keyword combinations to precisely define a topic to discover all relevant papers. Even for the smartest people, making connections and (re)discovering similar concepts can be difficult.
As long as this is the case, artificial intelligence cannot be trusted, and humans will need to carefully check all the content of artificial intelligence output after text mining. This is a tedious task and goes against the real purpose of using artificial intelligence. To solve this problem, we need to make scientific papers not only machine-readable, but also machine-readable.Understandable, By (re)writing them in a special type of programming language. In other words: teach science to the machine in a language that the machine understands.
Writing scientific knowledge in a programming-like language would be boring, but it is sustainable because new concepts will be added directly to the scientific library understood by machines. In addition, as machines are taught more scientific facts, they will be able to help scientists simplify their logical arguments; find errors, inconsistencies, plagiarism, and repetition; and highlight connections. Artificial intelligence that understands the laws of physics It is more powerful than AI that only trains on data, so science-savvy machines will be able to help future discoveries. Machines with rich scientific knowledge can help, not replace, human scientists.
Mathematicians have already started this translation process. They teach mathematics to computers by writing theorems and proofs in languages such as Lean. Lean is a proof assistant and programming language in which mathematical concepts can be introduced in the form of objects. Using known objects, Lean can reason about the truth or falsehood of a statement, thereby helping mathematicians verify proofs and determine where their logic is not rigorous enough. The more mathematics Lean knows, the more it can do.this Xina Project The goal of Imperial College London is to enter the entire undergraduate mathematics curriculum in Lean. One day, the proof assistant might help mathematicians conduct research by checking their reasoning and searching for the vast amount of mathematical knowledge they possess.