TikTok 的 AI 影片摘要功能火爆上線：名人變成藍莓等等！

TikTok, the popular video-sharing platform, recently introduced a feature called 'AI overviews' that uses artificial intelligence to summarize what is happening in videos. The goal was to help users quickly understand the content of a video by reading a short description generated by AI. However, this new feature did not work as planned. Many users noticed that the AI-generated summaries were often inaccurate, confusing, or just plain strange. For example, one video featuring the well-known TikTok dancer Charli D'Amelio was described by the AI as 'a collection of various blueberries with different toppings. ' In another case, a ballroom dance performance was summarized as 'a person repeatedly striking their head with a rubber chicken. ' These odd mistakes were widely shared and discussed online, with some users joking that TikTok must be playing a prank on them.

Because of the backlash and the many errors, TikTok decided to scale back the AI overviews feature. Now, instead of trying to describe everything in a video, the AI will only suggest products that are similar to those shown in the video. This means that if someone is showing a new gadget or a piece of clothing, the AI might help users find more information about those items. TikTok also allowed users to report any incorrect or bizarre AI summaries, so the company could learn from the mistakes and improve the technology. The company did not explain exactly why the AI made so many errors, but experts say that generative AI tools sometimes 'hallucinate,' or make up information that is not true.

TikTok is not the only tech company facing challenges with AI. Google and Apple have also had problems with their AI features, such as creating fake headlines or giving wrong answers. Even though AI technology is improving, it still sometimes makes mistakes that can be funny or even confusing. Tech companies are working hard to fix these issues and make AI more reliable. For now, TikTok's AI overviews will focus only on products in videos, and users can continue to give feedback if they see something wrong. This situation shows that while AI can be a useful tool, it still has a long way to go before it can fully understand and describe everything in videos without making errors.