So far I can recall other two instances of machine learning going unfortunately wrong: that time when a Google image algorithm tagged a black person as 'gorilla

So far I can recall other two instances of machine learning going unfortunately wrong: that time when a Google image algorithm tagged a black person as 'gorilla'; when recently, Google translate translated "a man has to clean" literally as "a woman has to clean" in Spanish. Should developers be now more aware of unintended consequences of this technology? Or is it too unpredictable? What can we learn from this examples?