Since OpenAI used ChatGPT to shock the outside world last fall, it has also made major technology companies begin to re-examine the AI track, and after more than half a year, a number of related products have also come to land. A few days ago, it was reported that TikTok is testing AI chatbots similar to ChatGPT in Japan and South Korea. It is reported that this chat robot is named Tako, has a cartoon octopus image, will appear in the upper right of the TikTok interface, users click on the icon will enter the relevant interactive interface.
Not only can Tako answer users' questions, but they can also initiate topics and adjust tone and style based on user feedback. In the demo released by TikTok, if the user said to Tako, "I'm bored", Tako responded, "Then I will recommend you a few fun videos" and push relevant content based on the videos that users have watched before.
In addition, Tako can also identify pictures or videos sent by users and make comments or recommendations based on them. For this feature, TikTok said that Tako aims to provide users with a richer and more personalized experience, and increase the interaction and stickiness between users and the platform.
It is worth mentioning that due to the relatively sensitive situation that TikTok is currently facing in overseas markets, the official also set up a disclaimer to tell users that Tako is an experimental chatbot, and the answer may be inaccurate. Because the company censors users' conversations with Tako for security purposes, it also warns users not to share personal privacy information with them.
It is not difficult to find from the relevant demonstration officially released by TikTok that although they are called AI chatbots, Tako and ChatGPT are obviously not the same thing at all, the former is more like a shrunken version of Bing Chat, and the core purpose is not to meet users' desire to communicate, but to guide users to watch more video content on TikTok. In a sense, Tako is more of an AI-enhanced in-app search tool.
After all, TikTok itself is positioned as an entertainment-oriented product, and its main selling point is "kill time". ChatGPT, on the other hand, is more productivity-oriented, so it's not suitable for TikTok.
In the past 3 years, TikTok has been like a traffic black hole, and it only took half the time for Facebook to exceed 1 billion users. According to data released by TikTok, TikTok currently has 1.601 billion users and 1.05 billion monthly active users. In addition, as overseas Internet manufacturers have come to their senses, almost all social platforms are trying to copy TikTok.
With Instagram's Reels, Snapchat's Spotlight, and Pinterest's Idea Pins coming out one after another, TikTok's previous booming era is over, so that's why TikTok is working on lifestyle community Lemon8 and video production software CapCut.
Therefore, in the eyes of many people in the industry, TikTok has now reached the stage where it needs to "practice internal strength" and let its 1 billion monthly active users play its due value. You know, a person familiar with the matter has previously revealed to the Wall Street Journal that TikTok's global advertising revenue target for this year is $12 billion. Therefore, the biggest significance of Tako may be to help TikTok revitalize content assets and make content discoverable by more users. The "magic" of TikTok is undoubtedly the recommendation algorithm, which allows users to continue to accept the baptism of content they are interested in, and then leave time for this APP.
Of course, the recommendation algorithm is not perfect, and under this information distribution model, users are passively accepting the content fed by the algorithm from beginning to end. So the problem is, the content selected by the algorithm must be time-sensitive and universal, which also leads to the algorithm shaping the trend at the same time, and the trend is also affecting the algorithm in turn. Therefore, the final result is that the content on TikTok is often time-sensitive and difficult to be used repeatedly, that is, the relevant content is precipitated as soon as the popularity passes. And these historical content hidden under the iceberg is not only difficult to provide the platform with the value of content consumption, but also becomes a burden that occupies server resources.
In order to ensure the supply of content, TikTok's practice today is to continue to find creators, but history has shown that those who have the ability to create and have sufficient desire to express themselves are always in the minority. In this way, revitalizing TikTok's existing content assets has become a more realistic option. This is actually the reason why Douyin, Kuaishou and WeChat in the domestic market have all done search business in the past two or three years, and search is an effective supplement and expansion of the recommendation algorithm, which can make the historical content precipitated in the platform also serve users.
Of course, search is a tool with a certain threshold. Nowadays, almost every platform is filled with a large number of homogeneous and one-sided low-information entropy content, search engines only rely on the search logic of "keywords" can no longer accurately filter the noise, and sometimes even can no longer play the value of information collection, such as Baidu is complained by users that it is not easy to use, because of the flood of low-quality information. At this time, the addition of AI makes users have a lower threshold for use, and the AI trained with the help of a huge corpus has better semantic understanding capabilities, so it can also better understand the problems and potential needs of users.
The most critical point is that AI also has self-learning capabilities, which can learn users' habits indifferently in their daily work and life, so as to continuously improve their understanding and search accuracy. The more land users use, the easier Tako will use, and eventually it will evolve into a user's "personal secretary" on TikTok. At the same time, in its continuous interaction with Tako, TikTok can also continuously understand the changes in users' needs and preferences to help it optimize its content distribution algorithms.
Perhaps in the future, the most likely landing scenario for generative AI is the content platform, after all, there are the most active users and the most demand.