Spotify’s generative AI “DJ” explained in excruciatingly boring detail
--
The much hyped AI “DJ” is now available in the US.
We took the time to review it.
Spotify’s “DJ” powered by generative AI and Sonantic’s voice to text technology serves up recommendations (like an old school terrestrial radio DJ) in six distinct ways, all of which have an underlying meaning.
In this post, we’ll break down the meaning of each mode:
1.) Based on recent listening
No surprise here, as “based on recent” listening is something that has existed on the mobile app for many years now. It’s based on your latest songs in rotation and how you engage (did you save, skip, or replay?) with each track. Based on your reaction, they’ll serve you more or less.
Nothing groundbreaking there.
2.) From your past
“From your past” is a new recommendation mode that’s clearly under the pantheon of the “based on” family. On first approximation, it’s based on your past listening history, but no details are available on how they quantify the past. A month? Six months? A year? It’s anyone’s best guess.
We’ll leave this one up for debate.
3.) Recommended for you
At first glance, “Recommended for you” is almost instinguishable from “based on recent listening”. However, there is a slight difference in the modes. “Recommended” for you seems to be based on recommendations coming from controversial technologies like “discovery mode”. Discovery mode currently serves up tracks through radio and autoplay mode, which we suspect is covertly being expanded to “DJ” as well.
4.) Throwbacks
“Throwbacks” is a brand new mode that seems to seize on the fact that most consumers (~ 70%) in the US choose to listen to catalogue (old music) versus frontline (new music). This mode makes a lot of sense given Spotify’s renewed focus on retention over flash in the pan engagement. Old music keeps people engaged and willing to explore new music.
5.) Editors’ picks
Perhaps the most dubious mode, “Editors picks” are recommendations currently highlighted by Spotify’s artist and marketing teams. They’re all the tracks that you’d find in playlists like “Rap Caviar” and “Today’s Top Hits”. We predict that labels like Sony, Warner, and UMG will be able to pay Spotify to recommend their frontline releases through AI DJ in the future.
(We suspect this is one of the “tools” UMG and Spotify collaborated on.)
6.) Trending music
Finally, “trending music” is a brand new mode that appears to cater to gen-z style records that are growing in popularity on Tiktok and Reels. Think JVKE’s “Golden Hour” or Dell Mac’s “So Sad” releases. Privately, many in the industry have theorized that Spotify has developed technology that allows them to accurately predict a hit before it becomes one. Whether or not that’s true, you’d find algorithmic recommendations like “trending music” in playlists like “big on the internet” and “my life is a movie”.
We fully expect Spotify to iterate on this product over time and give you the opportunity to pick your DJ’s voice (male or female), the number of skips (5 seems to be the magic number) between songs, and overall look and feel.
The implications for this product are clear.
Instead of hiring a radio promo guy to pay a mixshow DJ on terrestrial radio in a major market to spin your record at 3:00 AM, you can just pay Spotify cold hard cash to recommend it for you.
No human needed.
An unprecedented time to be in the music business.
+
Austin Staubus is the Managing Director of INR, a technology company located in Dallas. Since it’s inception in 2016, INR has been involved in many of music’s biggest streaming moments. Austin can be reached online at www.itsnorequests.com.
