For a long time, philosophers have pondered the purpose of language. Plato emphasized the importance of language, saying that thinking is “an internal conversation between the soul and oneself.” Modern scholars such as Chomsky have also argued that language is essential for thinking and reasoning.
However, recent research results in the neuroscience community contradict this. A representative example is Dr. Evelina Fedorenko, a cognitive neuroscientist at MIT. She scanned people's brains with fMRI and discovered something interesting. This means that the brain area that processes language is not activated during thinking or problem solving. “We couldn’t find any evidence that language was necessary for thinking,” says Dr. Fedorenko.
Research results in brain-injured patients also support this. It is said that even aphasic patients who lost their language skills had no difficulty solving math problems or playing chess. These studies show that language is not essential for thinking.
So what is the main purpose of language? Dr. Fedorenko firmly says ‘communication’. Research results support this, showing that words that are used more frequently are shorter in length, and that grammar rules tend to place words close in meaning. It seems as if language has evolved to optimize information transfer.
Now, shall we move on to the story of LLM, which has emerged as a big player in the artificial intelligence world these days? Thanks to learning from vast amounts of text, LLMs speak the language as fluently as humans. It seems as if it perfectly imitates the human ‘language network’.
However, their reasoning and problem-solving skills are lacking. Considering Dr. Fedorenko's theory that language and thinking are separate, this may be natural. You can say that LLM mastered the language tool well, but failed to develop the ability to think using that tool.
Of course, this study does not denigrate the value of language. As Professor Guy Dove of the University of Louisville says, “Language can be a tool to improve thinking.” When we think about democracy, we often think of conversations about democracy, right?
Language network: Specific brain regions were activated when participants performed language tasks. This area has remained in the same location over time.
Thinking network: Different brain regions were activated when the same participants solved puzzles or did different thinking tasks. But the linguistic network remained quiet.
People with aphasia: People whose language networks were damaged by brain damage were still able to perform tasks such as doing arithmetic or playing chess. This was further evidence that language is not essential for thinking.
The important thing is that language is not the only or essential tool for thinking. We can and have thought in much more fundamental ways. This seems self-evident, as babies struggle to solve problems and understand the world even before they learn language.
Watching the LLM makes me think again about what unique abilities only humans have. The power to go beyond simply exchanging information, but to creatively combine that information and create new knowledge. In order for humans and artificial intelligence to grow together in the future, won't we need both wings, 'language' and 'thinking'?
