Traveling with a large, active dog means forgoing air travel, and instead committing to road tripping, however long it might take. This year, as with the previous two Christmases, my wife, our dog, and I spent several days driving through the wilds of the Midwest to see family; this requires patience, fortitude, and a lot of music and podcasts. My wife does most (okay, all) of the driving, so her phone was linked into CarPlay, and she had selected several festive playlists for the trip.
Almost immediately upon hitting play, the music began skipping erratically. Songs cut out before their final, or even second, choruses. They faded into the next song but skipped its first minute or two. Some songs blipped in for a few bars and left. One of these was “All I Want for Christmas is You,” and I was mostly thankful for its truncation, but on the whole the situation was not conducive to singing along, and isn’t belting songs out-of-tune in the safety of one’s car the whole point?
I soon realized what had happened: AutoMix. Apple published the most recent release of iOS, 26, with this new feature in their Music app—and in fact, AutoMix is turned on upon upgrade without alerting the user. Apple touts the “enhancement” in a press release thusly:
Apple Music delivers an elevated listening experience with AutoMix, which mixes one song into the next, just like a DJ. Using AI to analyze audio features, it crafts unique transitions between songs with time stretching and beat matching to deliver continuous playback and an even more seamless listening experience.
What this means in the real world is that during play the application fades songs one into the other, attempting to sync up passages with similar tempos. This may require making the beat faster or slower and jumping over bits of the songs. Except it’s not bits, it’s often most of the track, and it’s not a “seamless listening experience,” it’s a freaking mess.
There’s a lot to hate about this feature. Apple wants users to think of it as having a virtual club DJ in one’s pocket, but it feels to me more like radio in the 60s and 70s, when announcers would cue one track into another and talk over the introductions and fadeouts in an effort to keep fickle listeners from turning the dial. Songs were treated like unvariegated slop instead of as discrete works with their own integrity. Producers during this time mixed tracks with an ear to radio play, and so almost every song ended in a tediously long fadeout; some even had fade-ins, or lengthy instrumental openings. Disk jockeys even had a term for talking over the entirety of a song’s opening, stopping just as the vocals began: “hitting the post.” Punk rock’s greatest gift to the world was to normalize songs that simply started and ended without any of that nonsense.
So there are aesthetic reasons to reject Apple’s feature. But even if you really, really want (for god knows what reason) your songs to fade in and out seamlessly, AutoMix ain’t it, kid. Its output is beyond broken: songs are truncated, lyrics are cut-off mid phrase. A feature this busted should never have been released, let alone be set as the default behavior. Somewhere along the line a coder, or a project leader, or QA tester, or even a clueless manager type should have listened to a tech demo and say, eww, no, what the hell is that, make it stop now.
But this is the problem with tech companies today: a fetish for novelty intersecting with the sunk cost fallacy. They spent x number of man-hours creating this electronic albatross and by god they’re going to ship it, even if (especially if) no one was asking for it in the first place. They have to be the first out of the gate, or someone else is going to corner the market in bad ideas. Apple’s press release brags that AutoMix uses AI in its analysis, and in many ways this feature is an encapsulation of the industry-wide fiasco that is Artificial Intelligence. AI is shoved on us whether we want it or not, whether it’s actually useful or appropriate. AI carries ethical and moral concerns. But most of all, it just doesn’t work.
Fortunately for my wife and dog and me, I was able to turn the feature off after a few minutes poking around the phone settings. Sweet relief! Then when we eventually arrived at my parents’ house, I got to experience the joy of turning off their TV’s motion smoothing. It was truly a magical Christmas. And so, as Tiny Tim observed, God Bless Us Every One!
Bonus post: More Eldritch AI Horrors
One of the stops on my round of Christmas Day socializing with family was to visit my older brother Rob and his extended family, and they are all perfectly wonderful people, but by that point my stupid autistic brain was pretty much shut down, so while everyone was playing party games the way normal people do, I was hyper-fixating on a YouTube video that was playing on the television, titled descriptively enough, “Bing Crosby, Frank Sinatra, Dean Martin, Gene Autry, Andy Williams, Nat King Cole🎄Golden Christmas.” I believe Rob was just looking for an easy playlist of seasonal crooner classics, and for the most part, the audio of this video fit the bill, providing familiar recordings that are probably not really public domain, but let’s not draw attention and pretend that they are.
But it wasn’t the songs that I was focused upon, it was the unearthly generative AI video that accompanies them. This consists of an eight and a half minute loop that depicts festive scenes of a holiday-decorated train traveling through impossibly white mountains covered in pine trees and snow. From time to time the train arrives in a Thomas Kinkade-esque village where happy people mill about shops that resemble dollhouses. Everything is rendered in the oversaturated glow of AI. And if one is prone to hyperfixation, strange things become apparent the more the loop repeats.
The number of cars in the train, for example, keeps changing. And the engine itself changes from a steam-powered to a cable car (although the cable car seems to have a vestigial smoke stack). The train’s riders stick their mittened hands out of the window and snow bursts off the knitwear in clouds, although the trees the passengers reach for are about 20 feet away. As the train streams through the town center, there are no platforms and the pedestrians wander by mere inches from their festive deaths.
Strangest of all is a happy couple consisting of a man in his 60s and a woman in her 20s (I guess AI is prone to midlife crises). The man carries a wrapped gift precariously pinched between the his right thumb and index finger. He stops and turns to his companion, smiling. As he lifts his left hand, a second gift magically appears, pulled whimsically from a parallel dimension.
Around the 1:23 mark, the real cosmic horror begins. At this point, the soundtrack also becomes generative AI, with a Dean Martinish voice singing unholy chimerical monstrosities that resemble actual songs, as one might remember them in a waking nightmare. I don’t know why the producer of this video decided that an hour and twenty three minutes was the limit for unauthorized song use, unless perhaps that producer were actually an AI as well. In any case, the soundtrack for the last forty minutes of the video is Christmas songs from the Upside-Down, as sung by a choir of Gremlins and Heat Misers. For me, the apex/nadir of the lyrics are:
so wave goodbye to Frosty,
as he melts away today
To which I say again, God Bless Us Every One!
