Professional Documents
Culture Documents
Tiktok'S Ai Strategy: Byte Dance'S Global Ambitions
Tiktok'S Ai Strategy: Byte Dance'S Global Ambitions
Assignment Questions
1. What role does AI play in Byte Dance’s TikTok product? Is it different from recommender
systems? How does AI create value for consumers of these products?
A Recommendation system uses AI to analyse the user activity to determine future
activity. For example, Netflix observes the kind of movies a user watches and determines
the next set of movies to recommend to the user. Similarly in TikTok,
For Content Consumers:
- When a user installs the app, he/she is presented with a set of videos without the need
to signup
- Depending on the users interaction with the app using metrics like click rate, reactions,
time spent on a video, reposts etc. the user preferences are built into a model
- The user is presented with videos of varying genres based on the location where
his/her activities are monitored and rated
- Basing on this the recommendation system for the user Is built personalizing the feed
for the user
For Content Creators:
- A creator signs up with TikTok in order to create videos
- On creation of a profile, if a user’s profile has adequate videos with regular uploads,
the algorithm tags them to be a clean account
- The content is first checked by the machine learning algorithms in place to prevent
insensitive content being propagated.
- Any tagged content is then sent to humans for further verification
- If a creator’s profile is tagged clean their videos are pushed into users feed creating
network effects.
- In case of creators who’ve reduced posting, their videos are also pushed highly in
order to increase engagement and ensure retention
TikTok’s AI algorithm works differently from other products in this space by limiting user
thinking. While all the other recommendation systems like Netflix and Instagram provided
choices TikTok differed by presenting videos directly without making the user spend time
on making choices.
3. What new application should Byte Dance build using its AI capabilities? What value does
AI provide there?
- In today’s world AI has a number of applications
- We can see some of the most popular applications of Byte Dance are its News App,
Gaming Platform and TikTok
- The pandemic has increased people’s interest in the stock market and trading in
general
- Byte Dance could build a trading application similar to Upstox, Zerodha etc.
- In the financial space there is a huge knowledge gap which prevents users from making
trades
- Byte Dance could leverage its Gaming Platform providing a gamified experience using
AI to simulate market conditions imparting knowledge while enabling users to make
trades
- It could also use its short form content to make interesting videos on various topics
based on the user’s profile and level of proficiency in the topic
- AI is important to personalize the gaming experience and to curate content based on
the user’s knowledge and depth of understanding
- This could help Byte Dance build a substantial reputation and break into the financial
work with synergies from its other applications
4. What should TikTok do in response to government fines? Can Byte Dance find a way to
censor its content to the satisfaction of Chinese government? What other solutions are
possible?
- With Byte Dance being a major threat to big companies like Baidu, Ant group, Tencent,
Meta etc. it is in a very risky situation
- The Geopolitical reputation of China is not helping Byte Dance’s stance
- It is important of Byte Dance to maintain its relations within China since majority of
Byte Dance’s revenues come from within China
Things to do to improve censorship of content :
- Make the policies transparent for people to view and understand the type of content
that is moderated
- Moderation at a global level might miss out on various subjects condemned by the
Chinese Govt.
- Instead, a localised approach with moderators with local moderators monitoring
content could be a better approach
- Content can be moderated at various levels:
o Removal of insensitive/restricted content
o Tagging certain content to give users a warning before viewing
o Blocking accounts of creators who repeatedly violate guidelines and policies
- It is also important to very closely moderate content that goes viral
- AI systems should be trained with sufficient content from all the restricted and
sensitive categories to enable efficient tagging
- The model needs to be constantly retrained with newer issues emerging expanding
the list of content that is restricted or frowned upon by the Chinese Govt
- Another important approach would be penalizing constant violators