Eric discusses the potential for a breakthrough in understanding human cognition through language models, suggesting that our current methods may be brute force and inefficient. The conversation delves into the idea that the true challenge lies in comprehending cognition itself, rather than just scaling language models.
