Steam has been experimenting, and nobody is paying attention

Steam remains one of the best services in the world, and I’ve always enjoyed their attempts to innovate. But it wasn’t until this week that I noticed they have a “Labs” division showing off experiments. The work they’re doing is rather fascinating.

You can visit Steam’s Labs by going to your main Steam interface and finding it next to a bunch of other buttons you probably never pay attention to. So here, I’ll show you in this handy image:

Once you click on it, you’ll see a list of experiments. There are only three so far, and this is the latest one: the “Automatic Show”…

The Automatic Show tries to combine several modest ideas into a crazy package. One idea is to convert every Steam game into a short video preview automatically by having four screens showing its trailers and gameplay simultaneously, with sound effects and music from the game. Another is to automatically categorize and compile these short videos into longer compilation videos revolving around a common aspect. For example, the “New and Popular games of September 9th, 2019” could be generated and compiled automatically without any humans involved.

In one example, we see a bunch of three second clips of horror games. This may seem silly at first, but I found myself appreciating the quick impression they offered, as well as the overall sense of what’s out there.

Here’s the crazy part: Valve wants to automate a generated text-to-speech description to go along with every game based on its reviews and description, and have this voice play over longer video clips as narrator or announcer. Sadly they don’t have the voice part figured out yet, so a Valve employee is reading what they computer has generated in the example video shown. Judging by what I heard, it’s actually okay.

Would you watch a computer-automated video game promotional show? To be honest, I think I would.

The Interactive Recommender

Remember that I said there have been three experiments already? One of the other ones is a “machine learning” recommendation service, which you can test out yourself.

As you can see from the above screenshot, I gave it a spin. I can confirm that it did a very good job of finding games that I was interested in. Of the forty or so games it recommended, I added about nine of them to my Wishlist already.

When you use the sliding bars near the top, you can look for more obscure or popular titles, which I found to be a really interesting choice for Valve to include. They must realize by now that popularity is not the only thing people want to see. A lot of people are hunting for the “hidden gems”.

You can also filter results and add keywords, as well as limit the time window of their release. It apparently looks at your purchase history, play history, and all the other data they’ve gotten about you in order to guess what you’re likely to buy and actually play. I think this is the big point people might miss if they don’t stop and think about the recommendations.

Like many of you, I have a giant backlog of Steam games that I’ll never play. This includes a lot of big blockbuster titles that were on sale and take a lot of effort to get into. In reality I tend to play small indie games with multiplayer features and quick action a lot because I get together with friends and play couch co-op with them through my Steam Link device. I also love to support smart little indie teams and promote their stuff. Perhaps this is why Steam’s algorithms didn’t recommend any big titles to me, but almost exclusively pixel art indie titles that I find charming.

Combine the Automated Show with this smart recommendation list and I think you might have an extremely compelling sales pitch.

Tagged , , , . Bookmark the permalink.

Comments are closed.

  • Archives