Steam Labs looks at users’ playtime and suggests what games they should buy

Valve just announced Steam Labs, which includes three AI features: micro trailers, interactive recommended game lists, and half-hour auto shows.

Valve has launched an experimental service called Steam Labs. In it, the service will use AI tools to scan the user’s playing time and recommend the games they should buy next. According to the official website, the service is used to “test the discovery of games, videos, machine learning and more.” Steam Labs currently has three experiments running: micro trailers, interactive recommended game lists, and half-hour auto shows.

Micro trailers allow Steam to recommend 6-second trailers by genre, select from curators, or other categories to you. If you are interested in a certain game, you can click on it to go to the store page. This feature is mainly for inspiration, as well as coordination with the account Steam Trailers in 6s on Twitter.

MicroTrailers is now an official part of Steam Labs. Check it out: https://t.co/f3lxDD2q4E

— Steam Trailers in 6s (@microtrailers) July 11, 2019

Show automatically is programmed like short game clips, but combines multiple micro trailers of each game to display to the music. Steam’s original goal was to have the description automatically read out from the game’s store page, but it seems Valve has struggled. Steam has admitted the voice is too dry.

Finally, we have the interactive offer list. It is a “neural network model used to recommend games based on user play history, among other data” based on “millions of Steam users and billions of plays.” In addition, you can adjust to see new or older titles, as well as limit games by tag.

The service comes after a barrage of criticism that Steam has faced following its disastrous annual Summer Sale, when it failed to deliver new games to users.