Hans Moravec wrote in 1988 that it is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers. He noted the difficulty lies in giving them the skills of a one-year-old when it comes to perception and mobility.
Skills that appear effortless to humans required millions of years of evolution to develop while abstract reasoning abilities like mathematics are evolutionarily recent additions to the human brain. The oldest human skills are largely unconscious and so appear to us to be effortless yet require complex biological processing.
Rodney Brooks decided to build intelligent machines with no cognition in the 1980s. He stated he would only build sensing and action capabilities and completely leave out what traditionally was thought of as artificial intelligence.
By the 2020s computers were hundreds of millions of times faster than in the 1970s due to Moore's law. Additional computer power finally allowed handling perception and sensory skills as Moravec predicted in 1976.
Andrew Ng presented a highly imperfect rule of thumb in 2017 stating almost anything a typical human can do with less than one second of mental thought can probably be automated now. Machine learning algorithms have begun bridging the gap between reasoning and perception capabilities.