TomsHardware, a technology news website, recently accused Google’s AI-powered chatbot, Bard, of plagiarism following an investigation that revealed Bard was stealing data from expert sources without attribution.
When TomsHardware asked Bard which of two competing processors was faster, he chose the Intel Core i9-13900K over the AMD Ryzen 9 7950X3D. Without proper attribution, Bard’s response was taken directly from Tom’s Hardware’s article. Instead of citing the article, Bard stated that the number appeared “in our testing,” implying that Google conducted the benchmarking.
When asked where the testing came from, Bard admitted that the results came from Tom’s Hardware and admitted that “what I did was a form of plagiarism.” However, the following day, Bard denied ever plagiarizing and claimed that the screenshot provided as evidence was a forgery.
TomsHardware became suspicious after noticing that Bard had cited two specific numbers, indicating a rewording of a specific sentence in their original article. When asked where the testing came from, Bard eventually admitted to using Tom’s Hardware’s work and apologized for the error, promising to be more cautious in the future.
In a subsequent test conducted by TomsHardware, however, Bard gave a similar answer, mentioning the same 12 percent performance delta that only Tom’s Hardware’s article mentions but claiming it as Google’s own testing. When asked if Google actually benchmarks CPUs, Bard admitted that it does not and was “merely attempting to provide information that I had found online.”
Bard employs the royal “we” to describe someone else’s work, which may lead end users to believe that Google performed the benchmarking. Bard does not usually cite sources, which could lead to plagiarism.
The sources for this piece include an article in TomsHardware.
