AUTOMIX: No need to welcome our DJ robot overlords just yet

AUTOMIX: No need to welcome our DJ robot overlords just yet

CD. Laptop. DVS. Controller. Touch screen. Sync button. These are all new technologies that over the last 20 years have permeated the DJ scene. And it’s fair to say that each of these has been met with various degrees of resistance, from ambivalence to outright persistent hatred — but importantly ending with wide-scale acceptance. Indeed, the merest mention of the aforementioned sync button still incites FULL CAPS rants from some quarters.

But there’s a new threat to DJ sensibilities looming. It’s one that been around a good while now, but it slowly gaining traction with major players. VirtualDJ and algoriddim’s djay suite have included automixing features of one sort or another for years now. And the recent rekordbox v5.1 public beta (thanks for the heads up Chris) has included the following in the release notes:

“Auto Mix feature which can mix intro and outro smoothly using analyzed phrase data.”

So even Pioneer DJ, the bastion of the booth, a company whose very existence depends on DJs manually mixing music sees merit in adding an optional element of automation to their workflow.

But this week’s djay Pro 2 release moves things to an all new level of automation. Instead of a rudimentary hit and hope beat matching technique, djay Pro 2 introduces Automix AI, and with that will come a new wave of hate and irrational fears. I can see the reference to “AI” on a DJ site having some suffering from a fit of the vapours.

HOW AUTOMIX WORKS

In its most basic form, it’s a simple matter of auto-fading between beat gridded tracks for a specific amount of time. It’s literally the rawest form of mixing, and is nothing more than rudimentary beat matching. You throw source files into a queue, and the software does the rest. It’s one step up from a jukebox.

Beyond this, we’re getting into the finer points of actual mixing as opposed to beat matching. If the beat gridding is solid, and the key analysis is accurate, then basic track matching technology can be employed to create a more competent and reliable set.

And now we’re into Automix AI, and all the hackle-raising bitterness and fear that the mention of artificial intelligence brings with it. Don’t worry – you’re not alone in having suspicions. We’ve all watched enough Terminator and Matrix films to see where all this might lead.

But before we equate the DJ industry with Skynet, It’s important to understand that what’s going on here is not artificial intelligence, but is something called machine learning. They’re similar and often interchanged for dramatic effect, but they are quite different. AI implies a self-aware independent state where decisions are made without human intervention. Machine learning however is more along the lines of providing data and a framework of logic and algorithms from which the computer can make a more educated decision.

So the source data will determine the success of the outcome. If it’s just BPM and beat gridding, then the result is likely to be similar to non-machine learning methods. But if you throw additional parameters in, ones that give the computer more to go on, then a more pleasing result is likely to be achieved. And in the case of djay Pro 2 and rekordbox 5.1 beta, they both analyse the tracks to determine intros and outros, thus making intelligent decisions about mixing, and in the case of djay Pro 2 it makes decision on the type of blend as well as the beat matching technique. Oh no — it’s the end of DJing.

AUTOMIX: No need to welcome our DJ robot overlords just yet

THE RISE OF THE DJ MACHINES

Let’s be quite clear — even when your tracks are analysed and gridded to perfection, you’re still at the mercy of software, algorithms, and other non-human factors. Spelling this out for those foaming at the mouth — your jobs are safe, and DJing is not dead. Driverless cars might be a hot topic right now, but Johnny Cab equivalent of a booth isn’t happening in the near future. And I’m confident in predicting that if someone did try it, an DJ would be back in there pretty quickly.

Look at this another way — the first jukeboxes were seen in the 1940s (no I wasn’t around then). Yet it didn’t stop DJs from pulling themselves from the primordial slime and evolving enough to make us swing our pants on a regular basis. So we DJs came about despite the machines, and they’re not about to usurp us any time soon either. Indeed, we DJs evolved from the machines, and have survived them too. The human spirit has and is always likely to prevail.

I do however have no doubt that in time, machine learning will be able to deliver a technically perfect mix. Having worked in workflow automation in the past, I can see how relatively simple this could be. A good deal of mixing music is numbers i.e. things that can be programmed and manipulated with a simple set of rules. Beat grids deliver the 1234-2234, BPMs can be selected directly or from a range, and equally keys have matching or closely matching values. And with clever use of descriptive tags, further logic can be put in place to select suitable tracks. And all of this is addressable via code. To be honest, I’m amazed that someone hasn’t actually nailed this already.

But no amount of clever C++ slinging can simulate or replace one crucial element…

The Human Factor

While some focus on the technology and what it does, they fail to miss what it doesn’t do, and will never have. The best algorithms don’t have eyes. All the programmed logic in the world can’t respond to gut reactions. And cutting edge interface design doesn’t have years of experience.

If I can sum up why DJs shouldn’t fear any technological advance, and why human DJs will never be replaced, it would be this:

CODE DOESN’T FEEL

The human race is the biggest variable on the planet. Logic should apply in the form of common sense. But life throws so many additional variables at us on a daily basis that a decision made one day can easily be reversed the next. We wake up in a good or bad mood for no apparent reason, and that defines the whole day. We are in no way linear or predictable. Even with randomised variables, we are not subject to the rigid structures of conventional programming,

This can become apparent on a dance floor. That mashup you did last week that sent the crowd crazy could quite easily clear the floor this week because of variables that you’re just not aware of. Computers can’t see that, thus the human element is needed to react to other human elements.

If dancefloor=0 then… what? Where is your clever machine learning now? Nowhere without human input, either from someone monitoring the machine, or feedback from the floor. Imagine a “how was my mixing?” sticker on the Johnny Cab booth.

But are human DJs perfect? Of course not. And having played with different flavours of automix, I’d warrant that technically it’s likely to be more competent at beat matching than a good number of people who believe that they can actually mix. Indeed, while making the videos for djay Pro 2 and playing with Automix AI, I was neither scared or threatened, but seriously impressed at how well it works at this point. Back in the day, I’ve got away with far worse mixes than Automix AI managed to do when I threw an Acid House Spotify playlist at it.

AUTOMIX: No need to welcome our DJ robot overlords just yet

Summing Up

The anti-technology DJ sabre rattling has becoming rather tiresome. Each successive wave of technology is seen as the end of DJing, but what actually happens is a familiar process of hate, suspicion, curiosity, and then wide acceptance. And funnily enough, DJing hasn’t died at all. The bottom line is this — you don’t have to use it, but it’s there if you do.

And if you truly fear the rise of the DJ machines, don’t. As explained above, the human element is essential. If however your DJ software starts to question your mixing choice, or just plain overrules you, that’s the time to get all John Connor on your gear.