by Nikolas Grafakos
Fermi “Machines” is an investigation exploring the idea of building machines to think broad instead of deep by analysing the process of how humans use broad knowledge to solve (fermi) problems that are designed with machine constraints.
Fermi problems are a common practice technique to help humans develop a broader way of thinking. They allow you to formulate an approximate answer to a problem based upon a sequence of questions and logical assumptions. Used in engineering to help people learn how to break big problems into manageable parts by asking the right questions, they ultimately build resilience.
Our human ability to think broadly has allowed us to thrive in Wicked learning environments, such as politics and ethics, by understanding how to solve complex problems. Unlike us, Artificial Intelligence has been limited by its performance to solve problems only in specialized and narrow fields, defined as Kind learning environments, such as games like chess or Go.
If we could understand how humans utilize their broad thinking to solve complex problems it may lead to expanding the capabilities of machine intelligence.