Moravec’s paradox is the observation by the robotics researcher Hans Moravec, published in 1988, that tasks which are easy for humans to perform, such as walking or facial gesture recognition, are difficult for machines to replicate; and conversely that tasks which are difficult for humans, such as mathematical calculations or large-scale data analysis, are relatively easy for machines to accomplish.[1]
Moravec went on to offer an explanation for his paradox in terms of human evolution:
The mathematician and artificial intelligence (AI) researcher Marvin Minsky had made a similar point two years earlier, suggesting that the most difficult human skills to reverse engineer are those below the level of conscious awareness: “In general, we’re least aware of what our minds do best … we’re more aware of simple processes that don’t work well than of complex ones that work flawlessly”.[4] The cognitive psychologist Steven Pinker wrote in 1994 that “the main lesson of thirty-five years of AI research is that the hard problems are easy and the easy problems are hard”.[5]
Although AI approaches such as large language models (LLMs) can capture large amounts of data and apply it to new problems, they can only be trained on communications between humans. So nowhere can any such model find instructions on how to move an arm to clean a greasy pan or walk, for instance, because every human already knows how to do that. Robots lack the basic skills humans take for granted, such as how to hold a pencil, or pick up a knife.[6]
