In his 1998 essay “When will computer hardware match the human brain?” (Journal of Evolution and Technology, Vol. 1), roboticist Hans Moravec estimated that matching overall human behavior would take about 100 million MIPS (million instructions per second). He arrived at the figure by calibrating against the vertebrate retina - which he reckoned does roughly 1,000 MIPS of work - and scaling up by the brain’s much larger size. He predicted the required hardware would become available “in cheap machines in the 2020s,” and refined the date to before 2030.