id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
lesk-fragility-2021		lesk-fragility-2021	2021-02-23	11	.pdf	application/pdf	4796	474	63	Machine learning systems have a set of data for training. of the real problem (if you train a machine translation program solely on engineering documents, there may be a lot of training data, including many noisy points, and the program may decide on Many popular magazines have discussed this problem; Forbes, for example, had an explanation of how the choice of datasets can produce a biased result without any deliberate attempt to used to suggest malicious creation of training data or examples of data designed to deceive machine learning systems. blood pressure, and lower blood pressure decreases the risk of heart attacks." Then I have to explain that the paper evaluates 32 possibilities (prior/current ownership ⇥ cats/dogs ⇥ 4 medical compare the performance of machine learning systems for medical diagnosis with actual doctors If a program is constantly learning from new data, there is no list of previously fixed failures to	./cache/lesk-fragility-2021.pdf	./txt/lesk-fragility-2021.txt