|
Briefing
|
- Game Recommending AI – Gaming company Valve announced new game recommendation engine based on neural network called Interactive Recommender that learns from games users play, rather than user-generated metadata
- Old Algorithm – Previous recommendation engine relied on metadata, such as user tags, user-curated lists, aggregate review scores and sales data, which developers used to optimize game discovery
- AI Training – Neural network learns about games as they are played by user, with Valve making it available on website for public trial by gamers and developers
- Player Analysis – AI compares number of hours users play in game with millions of other Steam users to recommend possible games user might like based on games played by other players with similar play habits
- User Parameters – Gamers can also select parameters for AI’s suggestions, such as popularity or niche, and older versus newer titles
- Complementary Algorithm – AI complements (rather than replaces) another Steam algorithm Discovery Queue, which is better for discovering new game releases, as Interactive Recommender relies on games users have already played, which can put new games at a disadvantage
|
|
Accelerator
|
|
|
Market Disruption
|
|
|
Business Model and Practices
Business Model
and Practices
|
|
|
Sector
|
Media and Entertainment
|
|
Function
|
Customer Experience and Service
|
|
Organization
|
Valve Corp.
|
|
Source
|
|
|
Original Publication Date
|
July 11, 2019
|
