14:00 on April 23rd,Yang Yiming, Professor at the School of Linguistic Sciences and Arts, Jiangsu Normal UniversityBeing a guest in the 367th issue of "New Apricot Forum"，Yu Zhixin Building Room A606 gave an academic lecture titled "The Difference between Generative Large Language Models and Human Natural Language Generation"。Lecture hosted by Professor Yue Lijing，Professor Wang Hui、Professor Kou Xin、Associate Professor Liu Wen、Associate Professor Zhang Yanfen and others attended，Attracted nearly a hundred teachers and students to listen。
The beginning of the lecture，Professor Yue Lijing briefly introduces Professor Yang Yiming’s research directions and main academic achievements，Speaking of Professor Yang Yiming’s extensive interests，Early research on phonology，The "perspective separation method" he founded is known as "a new method of modern sound research"。Mainly devoted to neurolinguistics research Bet365 gamesin recent years，is the main pioneer and founder of the field of neurolinguistics research in my country。
First，Professor Yang Yiming reviews the evolution of language research objects。He pointed out that every change lays a solid foundation for the birth and development of large language models。With the rapid advancement of technology，The development of large language models is also changing with each passing day。Professor Yang specifically pointed out，The parameter size of large language models has achieved significant growth，The continuous integration of new technologies makes the language model more powerful。At the same time,The concept of technology sharing and openness further promotes the widespread implementation of large language models in practical applications。He then systematically combed the development process of large language models，And divide it into three stages for detailed explanation，In-depth analysis of the functions and characteristics of large language models in processing language。
Next，Professor Yang Yiming conducted a series of in-depth experimental studies，Elaborate on the significant differences between large language models and human language。He mainly conducts comparative analysis from two aspects: bet365 games best slotslanguage acquisition mechanism and language processing mechanism。When exploring the mechanism of language acquisition，Professor Yang Yiming pointed out，Language learning of large language models，That is, its training process，is completely dependent on language data and languageCharacteristics，An imitation process similar to "parroting"。Relatively，Human bet365 games best slotslanguage acquisition has an innate genetic mechanism，Enables us to master and develop corresponding language abilities with only a small amount of language stimulation。At the level of language processing，The difference between large language models and humans is reflected in the input of information、Processing method、Multiple dimensions such as information understanding and content output。Professor Yang Yiming emphasized，Although large language models can imitate human language behavior，But due to the lack of innate biological mechanism of human language，It does not really understand the deeper meaning of language，Therefore, there is an essential difference from human language in terms of innovation。
Finally，Professor Yang Yiming is sure that large language models are important in text generation、Translation、Powerful features in areas such as programming and bet365 online betting appcreation，But also pointed out its existing problems。For example，The preference of big language models for big data leads to high resource consumption。At the same time,Because its training data mainly comes from the open Internet，Contains a large number of unverified、False or even wrong information，Even if filtering is performed during the data preprocessing stage，It is also difficult to ensure full compliance。In addition，The internal organizational structure of the large language model still needs further improvement，And it lacks a self-monitoring mechanism，Sometimes unrealistic "hallucinations" occur，Seriously talking nonsense。In response to these problems，Professor Yang Yiming proposed two improvement paths: First, promote the upgrading of AI technology based on language processing and language cognition；The second is to build a “language brain” based on the concept of “pre-installation”。These cutting-edge perspectives point the way for future development of large language models。
In interactive session，The on-site teacher and classmates conceptualized the psychological process、Ask Professor Yang Yiming about baby guidance language and other issues，Professor Yang Yiming answered everything bet365 games best slotspatiently and carefully。
Professor Yue Lijing expressed his gratitude to Professor Yang Yiming for his wonderful speech on behalf of the teachers and students of the School of Liberal Arts。Professor Yue Lijing pointed out，Professor Yang’s lecture not only broadened our horizons，Show us the vast space of linguistic research，Also remind us，Although large language models excel in some aspects，But there is still a lot of work for us humans to complete。Therefore，While we cherish and utilize the convenience brought by large language models，We should maintain independent thinking and creativity。The lecture ended successfully with warm applause。
