Otto Lappi

While “superhuman AI” routinely achieves tremendous performance in tasks which humans find challenging, it may seem paradoxical that the best available algorithms struggle on many tasks which humans find easy and natural. Especially in domains based on physical activity humans have a clear advantage: expert performance in any sport is beyond the dexterity of current machines. Even apparently simple everyday tasks like finding one’s way around a kitchen and making oneself a sandwich or a cup of coffee are vastly beyond current perception, planning and motor coordination algorithms. To identify what gives human intelligence this advantage in embodied skilled performance requires bringing together cognitive science, computer science/AI, neuroscience and even philosophy. Experimental research “in the wild” can provide data on human performance beyond narrow laboratory tasks (which are not a major challenge for machines). But such synthesizing findings into a rigorous framework for human-machine comparison requires engaging the fields of AI and robotics at a foundational level. This is Otto Lappi’s purpose in the Intelligence3 fellowship.