TikTok, Like Everyone Else, Is Testing Its Own AI Chatbot

TikTok is publicly testing an AI-powered chat companion in its app that won't only change how you watch content but also adds a knowledge-building aspect to those viral short videos. Social media platforms are no strangers to AI, as they rely on artificial intelligence for everything from serving tailored recommendations to moderation and even identity verification. With generative AI like ChatGPT quickly redefining how the internet works, social media apps are paying attention, as well. 

Snapchat recently put a ChatGPT-like bot in the app. Google's DeepMind has already deployed a visual language model called Flamingo that generates descriptions for YouTube shorts. It looks like TikTok is next. First spotted testing by Watchful, TikTok's answer to Snapchat's My AI bot is called Tako and it sits right at the top of the on-screen controls you see while watching vertical videos. TikTok has officially confirmed that Tako is being tested to a limited capacity in multiple markets, including the Philippines, but not in the US. 

While watching a video, users can tap on the Tako icon alongside the right edge of the screen to bring up a chat window. Here, users can ask the bot to provide relevant information about the contents of the video. For example, while watching a food video, they can ask Tako about the delicacy's history, recipe, and cultural importance, among other relevant queries.

TikTok is playing it safe with Tako

In addition to pulling relevant information that an internet-connected AI companion does, Tako is still very much a TikTok product. As such, it will help users find more videos of the kind they are already watching, if they ask the AI bot to do so. TikTok appears to be pushing Tako as a way to further fine-tune the videos you are recommended based on your interests, instead of scrolling past an endless feed of algorithmically-curated videos.

Notably, TikTok will keep the bot away from accounts belonging to minors, likely as a safety measure because AI-generated answers can occasionally be disturbing or downright inaccurate, and there's a healthy precedent for that out there. With those drawbacks in mind, TikTok warns users that Tako's answers might not always be accurate and that users shouldn't rely on it for medical, financial, or legal advice. According to TechCrunch, TikTok will let users delete their conversation history with Tako.

Interestingly, TikTok appears to be outsourcing Tako's AI tech to a third-party partner, instead of using its own in-house AI. It isn't hard to imagine why ByteDance would want to steer clear of OpenAI's GPT model, given the current state of trade tensions between the U.S. and China, with AI and semiconductors being at the center of the tussle. Notably, ByteDance already outsources its own AI to other companies and has recently started dabbling with its generative AI-based tricks. As for Tako, there is no word on its wide release.