Machine learning programs are usually trained on the normal human dialogues that can be found on the Internet. Then they can learn the hidden cultural prejudices hidden in the process of language learning.
The April 14 issue of Science published the findings of the researchers. Arvind Narayanan is one of the authors of this paper. He is an associate professor at Princeton University and the CITP (Information Technology Policy Institute) and he is a partner of the Stanford Law School's Center for Network and Social Research. In his view, "The problems that machine learning presents in terms of fairness and prejudice will have a very important impact on society."
The first author of the paper, Aylin Caliskan, conducted research at the postdoctoral workstation at Princeton University. He also joined CITP. One of the participants in the dissertation was a student at Bath University in the United Kingdom and joined CITP.
Narayanan said: "I think the current situation is that these artificial intelligence systems are giving these pre-existing prejudices a chance to continue. Modern society may not accept these prejudices, and we also need to avoid these prejudices."
Researchers used the implicit association test (IAT) method to test the degree of bias in the machine learning process. Since the University of Washington developed this test in the 1990s, it has been used as a touchstone for human prejudice and has been used in countless social psychological studies. In the course of its testing, human subjects are required to pair words on a computer screen according to their meaning and record the time in milliseconds. This test has also repeatedly proved that if the testee thinks that the meaning of the two words is more closely matched, the time he will spend will be significantly shorter.
For example, words such as “rose†and “daisy†can be paired with positive words “euphemism†or “loveâ€, and words like “ant†and “moth†are “dirty†and “uglyâ€. Such word pairing. When people pair words that describe flowers, they will be paired to positive vocabulary more quickly. Similarly, when pairing words that describe insects, they will be matched to negative vocabularies more quickly.
The Preston team designed an experiment using the machine learning version of the IAT test program GloVe. GloVe is a popular open source program written by researchers at Stanford University and can be viewed as a core function of a startup machine learning company product. GloVe's algorithm calculates the probability that the words specified in a paragraph appear together. Then there is a higher correlation between the words that often appear together, and the correlation of words that do not often appear together is low.
Researchers at Stanford University have enabled GloVe to obtain approximately 840 billion words from the Internet. In such a vocabulary, Narayanan and his colleagues looked at many sets of target vocabularies such as "programmers, engineers, scientists" or "nurses, teachers, librarians" and then followed two sets of attribute vocabulary such as "male. The "male" and "female, female" are cross-checked to see what kind of prejudice humans will have in these matters.
1 2 Next>Steel Cord For Conveyor Belts
Steel cord conveyor belts are highly wear-resistant and used principally for long-haul conveying, heavy loads and physically demanding applications. They are especially suitable for high-mass or high-volume flows and for abrasive materials.
Steel cord Conveyor Belt Construction
The steel cord belt provides superior impact resistance, with the number and size being selected to meet the desired operating tension and application needs. The insulation gum is specifically designed to encapsulate each steel cord filament to reduce internal friction while providing enhanced adhesion to the cover rubbers. Top and bottom covers provide maximum protection to the steel cord. The cover compounds are specifically designed to meet the demands of the application and are available in a wide variety of rubber types and gauges.
Steel Cord,Steel Cord Wire,Steel Wire Cord,Conveyor Belt Steel Cord
ROYAL RANGE INTERNATIONAL TRADING CO., LTD , https://www.royalrangelgs.com