ByteDance (including TikTok)

ByteDance, the owner of TikTok, is exploring the use of AI in the form of a chatbot similar to ChatGPT. This chatbot, codenamed "Grace," is being developed within ByteDance's enterprise collaboration platform, Feishu (also known as Lark). It is in its early stages and is currently undergoing internal testing with a select group of employees. ByteDance's venture into generative AI with this chatbot follows a broader trend in the tech industry, with other companies like Baidu and Alibaba also working on similar projects. (source)

ONLINE CHAT

ByteDance, the company behind TikTok, has introduced a new chat app called Feiliao (Flipchat) as an alternative to the dominant WeChat in China. Unlike WeChat, which covers daily communication, payments, and more, Feiliao focuses on interest-based social interactions. Users can create forums and chat groups centered around their hobbies and interests. The app's aim is to connect people with similar interests, fostering diverse and interesting interactions. Feiliao is part of ByteDance's strategy to diversify its app offerings and compete with Tencent, with ByteDance's apps accounting for an increasing share of users' time spent on giant apps in China. (TechCrunch | 2019).

CHATBOT

TikTok, the social media platform owned by ByteDance, is currently testing an AI chatbot named Tako. This chatbot, accessible through an icon on the app's interface, engages users in conversations, offers answers to inquiries, and provides content suggestions based on preferences. However, users should be cautious as the data inputted into Tako is shared with third-party service providers for generating responses. Although experimental, Tako has the potential to gather precise user data, enabling TikTok to deliver personalized content. The chatbot is restricted to users over 18 and its AI-generated responses may not always be accurate. This reflects TikTok's drive to incorporate AI-driven features for enhanced user engagement. (Source: Gulf Business, The Verge | 2023)

SELF-DRIVING CARS:

TikTok owner ByteDance, start-up Haomo.AI launch new infrastructure to boost autonomous driving push in world’s biggest car market (Source | 2023)

Haomo.AI and Volcano Engine, a cloud service owned by ByteDance, have established China’s largest dedicated computing centre for autonomous driving
The venture shows increased optimism about the technology in China, which is forecast to become the world’s largest autonomous vehicle market by 2035

Policies, or Practices

ByteDance values AI ethics and fairness. They work on reducing biases in content recommendations and invest in AI research for safety and user well-being.

The AI Behind TikTok's Addictive Algorithm: A Simple ExplanationToday we are going to explore the secret behind the addictive nature of social media apps that keep users glued to their screens, scrolling through seemingly endless content. Social media apps use a variety of AI systems to deliver a personalized and engaging experience to their users.

TikTok AI Live Trend: What Is It and Why Is It So Popular? - CloudbookletDiscover the fascinating world of TikTok AI Live Trend! Unravel the secrets behind its popularity and learn how to use this feature to create stunning videos with artificial intelligence filters.

HOW AI IS TRAINED

Bytedance trains AI for content recommendation and moderation on TikTok.

Data Collection: Bytedance collects user-generated videos and engagement data.
Data Preprocessing: Videos are preprocessed by categorizing content, detecting copyrighted material, and analyzing user interactions.
Model Architecture: Bytedance employs deep learning models, such as convolutional neural networks (CNNs) and transformers, for content recommendation.

Training:

User interactions and video data are used to train models.
Loss functions are used to optimize video recommendations and content moderation.
Models are trained over multiple iterations.

Evaluation: Bytedance assesses AI performance using metrics like user engagement, video completion rates, and content removal accuracy.

Fine-Tuning: Bytedance fine-tunes AI models by adjusting content recommendation thresholds and moderation policies.

Testing: Bytedance conducts tests to evaluate the impact of AI changes on user video discovery and content safety.

ISSUES

Data Quality and Quantity: Bytedance faces data quality issues when optimizing content recommendations on TikTok to enhance user engagement.
Bias and Fairness: Bytedance works on reducing bias in TikTok's content recommendation to promote diverse content.
Overfitting and Underfitting: Bytedance aims to mitigate overfitting and underfitting in TikTok's AI for content filtering and recommendation.
Interpretability and Explainability: Bytedance aims to enhance explainability in TikTok's AI for content moderation.
Scalability and Performance: Bytedance optimizes TikTok's AI for scalability during high-demand periods to maintain smooth content delivery.
Ethical and Legal Considerations: Bytedance works on ethical AI usage in content moderation and user data handling on TikTok.