王士元先生在第15届演化语言学国际研讨会开幕式的发言

Dear Friends,

I deeply regret not being able to join you at CIEL-15.  I have been very much looking forward to seeing old friends of many decades, & to making new friends.  Unfortunately, I am not in a fit condition to make an effective presentation, and I certainly do not want to pass on to you the virus that’s making all the trouble. 

亲爱的朋友们：

很遗憾没能参加CIEL-15。我一直非常期待见到几十年的老朋友，并结交新朋友。不幸的是，我因抱恙无法做个有效的演讲，我当然也不想把制造麻烦的恼人病毒传给你们。

However, as my title “Whither sapiens?” shows, I have been hoping to discuss with you a set of concerns that has been weighing on my mind for some time now.  Humans have evolved to become “Masters of the Planet” largely because we invented language, a mental technology which enables the pooling together and integration of knowledge across time and space. With the remarkable advances that came with the scientific revolution, humans are now a super-organism linked over 7 billion minds, all connected together electromagnetically via the Internet. 

然而，正如我的标题“智人何去何从？”所示，我一直希望与你们讨论一系列已困扰我一段时间的问题。人类之所以演化成为“地球主宰”，很大程度上是因为我们发明了语言，这是一种能够跨越时间和空间、汇集和整合知识的心理技术。随着科学革命的显著进步，人类现在是一个超级有机体，连接着70多亿个头脑，这些脑海里的思想都经由电磁的方法通过互联网连接在一起。

The most remarkable and surprising advances came during recent decades, when machines successfully hacked human language, and can now access all the knowledge accumulated over the millennia of human civilization. Electronics can process information and reproduce information many orders of magnitude faster and more precisely than biological brains. Defeating humans easily in board games is but a tiny foretaste of the much more powerful intelligence machines are developing. Solving intractable problems of protein folding in biology is a more dramatic example of the major benefits from this new intelligence. 

最近几十年来，最显著和令人惊讶的进步出现了，因为机器成功地入侵了人类语言，现在可以取得数千年人类文明积累的所有知识。电子可以用数倍于生物大脑的速度更快、更精确地处理信息和再现信息。在棋盘游戏中轻松击败人类，只是更强大的智能机器正在发展的一个小小的预兆。再举个更引人注目的例子，这种新智能的主要好处是可以解决生物学中棘手的蛋白质折叠问题。

Faced with these fundamental changes in evolution, how much longer will humans remain masters of the planet is now seriously debated.  Putting such cosmic issues aside, new and exciting questions arise for linguists.  How do the large language models (LLMs) acquire language differently from human acquisition?  How is language represented in LLMs as opposed to how it is represented in the brain, the so-called “Model brain alignment problem”?   If a degree of alignment can be found, can this knowledge be used to better understand the nature of language and cognitive disorders and to devise their intervention?

面对演化过程中的这些根本变化，人类作为地球主宰的时间还能持续多久，这一问题现在正在激烈争论之中。抛开这些宇宙问题不谈，语言学家们面临着新的、令人兴奋的问题。大型语言模型（LLM）习得语言的方式，如何与人类习得语言的方式不同？相对于语言在大脑中的表征方式，语言在LLM中是如何表征的，即所谓的“模型大脑对齐问题”？如果能够找到一定程度的对齐，这些知识能否用于更好地理解语言和认知障碍的性质，并制定干预措施？

I believe CIEL, with its mission to study evolution, is an appropriate forum to discuss these questions and stimulate research on them, especially by young friends with the courage to rise to the challenge.  Given my absence, I have invited Professor SHI Feng, one of the founders of CIEL, to voice these questions on my behalf.  I would of course be more than happy to discuss them at much greater detail at any time.

我相信，以研究演化为使命的CIEL，是一个合适的论坛来讨论这些课题并激发相应的研究，尤其是那些有勇气迎接挑战的年轻朋友。由于我的缺席，我邀请了CIEL创始人之一石锋教授代表我提出这些问题。当然，我非常乐意在任何时候与各位更详细地讨论这些问题。

            shiyuan, 2024/07/10.

             wsywang@polyu.edu.hk