Large language models (LLMs) like ChatGPT have become incredibly sophisticated, capable of writing poems, debugging code, and even passing standardized tests. But does this mean they think like us? New research dives deep into the cognitive patterns emerging in LLMs, comparing their decision-making, reasoning, and creative abilities to those of humans. The results are surprisingly complex. While LLMs demonstrate a remarkable ability to mimic certain human cognitive biases in decision-making (like the framing effect and anchoring bias), they also seem to lack some biases common in humans. This discrepancy isn’t just a quirk of AI; it offers a fascinating lens through which to examine the human biases embedded in the massive datasets these models learn from. When it comes to reasoning, the latest LLMs show a shift towards more deliberative, human-like thought processes, excelling in analogical reasoning and often outperforming humans. However, they can still struggle with deductive reasoning and fall prey to “hallucinations,” generating incorrect information with unwavering confidence. The research also reveals a curious dichotomy in creativity. LLMs shine in language-based creative tasks like storytelling but struggle with divergent thinking, likely due to their lack of real-world, embodied experience. Intriguingly, LLMs excel as creative collaborators, augmenting human innovation in diverse fields. This research underscores the importance of understanding not only what LLMs can do, but how they do it. Future research needs to address key cognitive areas like memory, attention, and knowledge representation to fully grasp the potential and limitations of these increasingly powerful tools. As LLMs become more integrated into our lives, understanding these nuances will be crucial for responsible development and application.