This post discusses recent research that shows a correlation between large language model (LLM) representations and human brain activity during language processing. The findings indicate that a linear transformation of embeddings from speech encoders aligns closely with neural patterns in the brain's speech areas. Similarly, embeddings from language decoders correlate with activity in the brain's language processing regions. This evidence suggests that LLMs may possess cognitive modeling capabilities reminiscent of human brain function, potentially countering the notion that they merely function as 'stochastic parrots'. The research is expected to pave the way for further discoveries regarding brain function and understanding cognitive processes.