For the 31)of learning, it needs large quantities of data,and the easiest way to find that data is to feed t with text from the 32)But these data contain some extremelybiased language. A Stanford study found that web-based Al connected white names with 33)words like "love,and black names with negative words like "failure" and "cancer "Luminoso Chief Science Officer Rob Speer is in charge of the data set ConceptNet Numberbatch, which is used as aknowledge base for Al systems. He tested one of Numberbatch's data sources and found obvious problems with their wordconnections. When fed the question "Man is to woman as shopkeeper is to.." the system filled in "housewife." So Speer34)for the clearing up of the biases in Concept Net. He 35)inappropriate connections and 36)them to zero, while maintaining appropriate connections like "man/uncle" and "woman/aunt." He did the samewith words related to race and religionBias 37)in human thinking. To fight human bias, it takes a human. So Al developers have a hugeresponsibiity tofind the problems in their Al and 38)them.

A)abandonedB)adjustedC)driftsD)fruitfulE)identifiedF)installG)intellectualH) internetI)originatesJ)perspectiveK)posiiveL)retainM)sakeN)stroveO)tackleThese days everyone is talking about the fascinating things artificial intelligence can do ike the human brain.Nevertheless,t might be wise to put府into 29)True it can deal with data more quickly and accurately thanhumans, but it can also 30) our biases (偏见). For the 31)of learning, it needs large quantities of data,and the easiest way to find that data is to feed t with text from the 32)But these data contain some extremelybiased language. A Stanford study found that web-based Al connected white names with 33)words like "love,and black names with negative words like "failure" and "cancer "Luminoso Chief Science Officer Rob Speer is in charge of the data set ConceptNet Numberbatch, which is used as aknowledge base for Al systems. He tested one of Numberbatch's data sources and found obvious problems with their wordconnections. When fed the question "Man is to woman as shopkeeper is to.." the system filled in "housewife." So Speer34)for the clearing up of the biases in Concept Net. He 35)inappropriate connections and 36)them to zero, while maintaining appropriate connections like "man/uncle" and "woman/aunt." He did the samewith words related to race and religionBias 37)in human thinking. To fight human bias, it takes a human. So Al developers have a hugeresponsibiity tofind the problems in their Al and 38)them.