The Original Version The following are some examples of how to use this story The following appeared: Quanta Magazine.
Which of the many abilities humans have is unique to them? The ability to speak has always been considered a unique human characteristic, at least ever since Aristotle said that humans were born with the language. “the animal that has language.” Researchers want to find out if certain aspects of the human language are unique and cannot be found in other animal or artificially intelligent communication systems.
The extent to which models of language can be used as a tool for reasoning about the language has been explored by researchers. Some members of the language community believe that language models can be used to reason about languages themselves. don’t Have reasoning skills, you can can’t. Noamchomsky (a prominent linguist) and two of his coauthors summarized their view when they published it in 2023. wrote in The New York Times You can also find out more about us here. “the correct explanations of language are complicated and cannot be learned just by marinating in big data.” These researchers argue that AI models are adept at using languages, but not able to analyze language in an advanced way.
In a recent article, this view was disproved. paper By: Gašper BegušA linguist from the University of California at Berkeley Maksymilian DąbkowskiHe recently earned his Ph.D. in Linguistics from Berkeley. Ryan Rhodes Rutgers University. The researchers put a number of large language models, or LLMs, through a gamut of linguistic tests—including, in one case, having the LLM generalize the rules of a made-up language. One of the LLMs was able to perform linguistic analysis in a way humans can’t. It was able to analyze language in much the same way a graduate student in linguistics would—diagramming sentences, resolving multiple ambiguous meanings, and making use of complicated linguistic features such as recursion. This finding, Beguš said, “challenges our understanding of what AI can do.”
It is timely that this work be published. “very important,” The following are some of the ways to get in touch with each other Tom McCoyThe research was conducted by, an unrelated computational linguist who works at Yale University. “As society becomes more dependent on this technology, it’s increasingly important to understand where it can succeed and where it can fail.” The ideal method to evaluate the ability of these models to reason is through linguistic analysis.
Complexity is infinite
It can be difficult to ensure that language models don’t know all the answers before putting them through a rigorous test. These systems are typically trained on huge amounts of written information—not just the bulk of the internet, in dozens if not hundreds of languages, but also things like linguistics textbooks. In theory, the models can simply regurgitate and memorize information they have been taught during training.
To avoid this, Beguš and his colleagues created a linguistic test in four parts. In three of the four sections, the model was asked to analyse specially constructed sentences by using tree diagrams. These were introduced first in Chomsky’s 1957 landmark book. Syntactic structures. The diagrams are used to break down sentences into noun and verb phrases, and further divide them into adjectives, adjectivals, prepositions and conjunctions.
One part of the test focused on recursion—the ability to embed phrases within phrases. “The sky is blue” This is a very simple sentence in English. “Jane said that the sky is blue” The original sentence is embedded in a more complex one. This process can continue indefinitely. “Maria wondered if Sam knew that Omar heard that Jane said that the sky is blue” This is also an awkward but grammatically sound recursive statement.
