Does AI understand human language?

The rapid advancement of AI, especially in language processing, has led many to speculate whether AI is on the verge of developing into AGI (Artificial General Intelligence), which possesses human-like, all-encompassing intelligence. This development has also sparked concerns about the potential dangers AI could pose to humanity. Within three months of ChatGPT’s public release, several notable individuals and organizations issued statements warning about these AI threats (March 22, 2023, Future of Life Institute [1]; May 1, 2023, Geoffrey Hinton [2]; May 22, 2023, OpenAI [3]; May 30, 2023, a group of AI CEOs and researchers [4]). However, there is also a faction of AI researchers who believe that achieving human-like general intelligence with AI remains a challenging feat.

At the Beneficial AI 2017 seminar, numerous AI researchers gathered to discuss AI. One major topic of debate was ASI (Artificial Superintelligence) [5]. The majority believe that it is indeed possible: once computers achieve AGI (Artificial General Intelligence), they will eventually develop into ASI. However, the time it will take to progress from AGI to ASI is uncertain and could be very short. The difficult part is reaching the AGI level. Most believe that this will take at least another 20 years, with some thinking it might not happen within our lifetime [6]. A significant obstacle often discussed is language…