Do you think modern LLMs are too focused on processing language at scale? Moreover, in the process, they are losing the ability to distinguish what information is at the crux of the question; especially when interpreting local dialects and linguistic nuances?

I firmly believe modern Large Language Models (LLMs) sometimes prioritize processing language over picking up what truly matters. In my own personal opinion, I would emphasize that language is not merely about words: context, culture, and intention and every language’s own humour and expression play a pivotal role in shaping it. Although their responses may be grammatically correct, they can miss the speaker’s actual meaning. Therefore, the challenge is not simply processing more data but understanding it meaningfully. LLMs must develop greater contextual awareness and respect linguistic diversity to communicate effectively across different communities. This has been brilliantly taken up by some local LLMs developed to tackle problems at their root.