The recent research in the field of generative AI and large language models (LLMs) has been marked by a shift towards understanding and emulating human cognitive processes more deeply. A significant trend is the exploration of cognitive abilities beyond mere language processing, such as metacognitive monitoring and visuospatial reasoning. Studies are increasingly focusing on benchmarks that mirror human developmental trajectories, such as the Clock Drawing Test, to assess AI's cognitive capabilities. These benchmarks reveal both strengths and weaknesses in AI models, highlighting areas where further development is needed to achieve human-like cognitive functions. Additionally, there is a growing concern about the internal validity of AI benchmarks, with research indicating that models might be exploiting superficial patterns rather than demonstrating genuine understanding. This suggests a need for more robust evaluation strategies to ensure that AI performance accurately reflects its capabilities. The field is also witnessing advancements in aligning AI models with human psychometric data, although this often comes at the cost of downstream task performance. Overall, the direction of research is towards creating AI systems that not only process language effectively but also exhibit cognitive abilities that are more closely aligned with human intelligence.