听说，这次来了很多业内大佬？
听说，这次有很多国内外学者？
听说，这是一次交叉学科的盛宴？
还听说，这次的学科论坛形式丰富？
当你带着新鲜的思考踏入这里，聚光灯将为你闪烁，自由的火花将在这里碰撞，让我们一起探索交叉学科的奥秘吧。
离本期交叉学科论坛启动就剩一周左右的时间啦！本次论坛可谓是大佬云集，干货满满！接下来就由小编带大家提前了解出席本次论坛的各位专家和报告内容：
James Lantolf教授
北京语言大学语言认知科学学科创新引智基地特聘教授
美国宾州州立大学荣休教授
主要研究领域：
社会文化理论（SCT），课堂二语发展
报告题目：
A Unified Theory of SLA: Possibility of Pipedream?
报告摘要：
A reasonable number of legitimate SLA theories is between 10 and 15, based on recent publications. Some scholars celebrate the fact that the field has so many theories and do not feel any need to reduce the number. But at some point it does become necessary for any field to reduce the number of theories it can tolerate because it is important for progress to determine which theories are viable and which are not. Many SLA researchers believe that theories and affiliated research have implications for language pedagogy.
Several solutions have been proposed to deal with the array of theories. One option has been to synthesize the claims and findings emanating from the various theories. Another has been to propose a set of criteria that theories must meet to be considered viable. Yet another is to try to find common ground among the theories. A proposal called in other sciences “adversarial collaborative research” (Melloni, et al., 2021) has special significance to compare the theories. In this presentation I will explain the implications of this proposal and will briefly discuss the nascent stage of a project that seeks to conduct adversarial collaborative research comparing the sociogenetic claim of Sociocultural Theory with the cognitively-based claim of Processability Theory.
Pedro Paz-Alonso研究员
西班牙巴斯克认知、大脑和语言中心教授以及语言和记忆控制研究小组主任
主要研究领域：
认知神经科学，语言系统，记忆系统
报告题目：
Thalamic involvement in language systems using advanced MRI protocols
报告摘要：
There is a strong need in the human neuroimaging field to implement reproducible research methods. Ideally, researchers should be able to replicate experimental data acquisition and reproduce the computational analyses presented in neuroimaging publications. Current scientific practices increasingly promote data sharing among researchers and labs. This facilitates colleagues in the scientific community and clinical practitioners to reproduce, verify, and further explore datasets and computational analyses. Moreover, these practices are evolving alongside continuous and substantial increases in algorithm complexity in functional and structural MRI methods. Over the last decade, our research group has developed MRI protocols for several purposes, such as identifying and segmenting thalamic nuclei at the individual-subject level, obtaining and measuring first-order relay human thalamic white-matter tracts, investigating the white-matter pathways between the human mediodorsal thalamic nucleus and the prefrontal cortex, and understanding the relationship between thalamic nuclei and their white-matter projections across the lifespan. Here, we will provide an overview of these neuroimaging protocols and their results, in particular on characterizing the involvement of the sensorimotor thalamic nuclei in human reading, speech comprehension and speech production language systems. Findings will be discussed in relation to current theories on subcortical contributions to cognition and on the role of thalamocortical interactions in language function.
江铭虎教授
清华大学人文学院计算语言学教授、博士生导师
主要研究领域：
自然语言处理、语言认知
报告题目：
从语言认知到大语言模型
报告摘要：
语言是人类区别于动物的最重要的特征，使人类的知识得以积累，创造出当今先进的科技和文明的社会。大语言模型在模拟人类语言能力方面已经取得了令人瞩目的成就，并且为我们探索人脑的语言认知机制提供了全新的视角。语言与人脑共同进化，其中脑容量起了决定性作用，而大语言模型只有在模型的深度和宽度足够庞大，通过学习超大规模数据的统计规律来构建对世界的深层次理解，其理解能力是通过能够整合多源上下文信息的Transformer框架来生成连贯和逻辑一致、且富有创造性的文本，标志着深度学习在AI领域的崛起。人类习得语言是通过多模态感官的结构化社会互动产生的，能够为许多词汇提供具体的外延含义；相反大语言模型必须从包含语言的单一信息流中获取世界知识，而不是将语言信息与外部经验联系起来。因此，大语言模型的预训练数据只有超出人类习得语言的几个数量级才能达到人类语言的水平。相反，人脑的神经元突触的数量比目前最大规模的大语言模型的权值参数大几个数量级，这些因素为我们理解人类语言的复杂性和未来的大语言模型发展提供了宝贵的启示。
更多专家信息，敬请期待来自引智基地的一线报道……
百问百答专栏
你有想对专家说的话吗？有想要和专家交流的问题吗？
是否对本次论坛的活动有疑问呢？
或者有哪些突然闪现的小tips.....
“百问百答”专栏将限时开通，收集来自您的奇思妙想，欢迎大家积极投递，我们期待与您的思想碰撞......
邮箱：clcb@blcu.edu.cn
