A new hafnium-based memristor inspired by the human brain could slash AI energy consumption by 70%. Researchers from Cambridge developed a stable artificial synapse that processes and stores data in the same place, using switching currents a million times lower than conventional devices.