TapTechNews May 15th News, at today's Google I/O conference, Google aggressively promoted its artificial intelligence named Gemini will take over the search engine field. However, in its flashy demonstration session, Gemini once again exposed the fatal flaw that has plagued all large language models so far: confidently giving wrong answers.
In the demonstration segment of Search in the Gemini Era, Google showcased its video search function, allowing users to take short videos and search by voice. The example in the demonstration was a stuck film camera winder. When the user asked why the winder can't move to the end, Gemini recognized the problem and provided repair advice.
However, the problem is that the answer provided by Gemini is overall laughable. One of the highlighted suggestions in the demonstration was to open the back cover and carefully remove the film, which may be the worst operation a user can do in such a situation (opening the back cover in non-dark environment will expose the film, damaging the already taken photos).
TapTechNews noticed that this is the second time Google has made a major mistake in demonstrating its artificial intelligence products. Last year, its Bard chatbot confidently claimed that the James Webb Space Telescope was the first telescope to capture exoplanets outside the solar system, which is not true.