Scientists from the University of Chicago (USA) conducted a study in which they showed how to use AI to write complex fake reviews. Such reviews can not be distinguished from real ones by modern methods, and unsuspecting readers find them highly reliable.
Reviews about restaurants were generated with the help of recurrent neural networks (deep training techniques), which previously trained on thousands of real reviews available on the Web.
According to the researchers, the generated responses were virtually indistinguishable from the real ones. So, the authors of the work showed that users not only did not recognize fake reviews, but also considered them as useful as real ones written by people.
The last, probably, guards most of all. Since this essentially means that the testimonials written by AI perform their main function – they purposefully influence people’s opinion.
It is noted that in such reviews, plagiarism was rarely found (with the help of software). This is due to the fact that the AI generated them knowingly, and did not pull out the words from these reviews.
Today there is a fairly large underground industry for writing fake reviews by people (for money). However, as Professor of the University of Chicago Ben Zhao (Ben Y. Zhao) in an interview with Business Insider, the implementation of AI can undermine it. The scientist says that he does not yet know about the use of such algorithms in this industry. But there is no guarantee that someone will not come up with something similar and will not use for mercenary purposes.
At the same time, researchers write that the responses of the neural network were still not ideal. It turned out that the algorithm used a smaller set of characters – and it was not difficult to notice. However, according to the authors of the work, future neural networks can be even more complex, and, accordingly, the responses generated by them will be more difficult to detect.
Jao notes that it’s not just about fake reviews about restaurants: such technologies can generally shake our beliefs about what is real and what is not.