Now it is 2019, Although AI has been applied in many aspects and provided a lot of convenience for our work, life and study, it has also created a lot of trouble, so that some people always joke that artificial intelligence is "artificial mental retardation." Recently, Yelp, the largest review site in the United States, released a hilarious version update note:
"We're deeply sorry for the program problems that occurred this week. We tried to train a neural network to eliminate all bugs, but it chose to delete all content... We had no choice but to roll back the version, but at least we have a 100% bug-free experience."
I wonder if this is what the AI thought after deleting the database
Here’s what happened:
Originally, Yelp intended to use deep learning to eliminate existing bugs and optimize and improve user experience, but the actual effect was ridiculous. The “naughty” neural network deleted all the content. As a last resort, Yelp could only roll back the version.
Netizens also made some interesting comments about the Yelp incident.
Apart from this blunder, as the largest review site in the United States, Yelp has been actively using deep learning-related algorithms in specific business scenarios, such as rating food photos and classifying images. In 2017, Yelp also made its internal data set public. In academia, the identification of fake reviews on online review sites such as Yelp is also a concern for deep learning researchers. According to Mika Juuti, a doctoral student at Aalto University, algorithm-based fake reviews are now easy, accurate and fast to generate. Most of the time, people can't tell the difference between real and machine-generated fake reviews. Based on this, Juuti proposed a solution in a study last year (2018): He and related researchers designed a classifier to identify fake reviews, which performed well, especially in the most difficult cases for human assessors to judge whether the reviews are true. However, there has been no relevant report tracking whether Yelp has put it into actual application to help users filter review information.
Referring to similar companies such as Meituan Dianping in China, the deep learning applications of online review websites are still mainly based on business scenarios such as text analysis, semantic matching, search engine recommendation ranking, and image quality ranking.
It is rare for Yelp to use real data sets to delete bugs but eventually cause all data to be deleted. As for how the neural network performed in the test environment, Yelp did not explain in detail. Last year (2018), many companies launched AI automatic bug testing tools, such as Facebook's SapFix. No matter how good the test data released by each company is, at least from this case, it is not as easy as imagined for AI to help test engineers detect and debug bugs.
提升卡
变色卡
千斤顶
京公网安备 11010802033920号
