“The Moravec paradox, which states that high-level reasoning is easy for AI while low-level sensorimotor skills are hard, still applies to current Large Language Models.”