ADC 2018 Lightning Talk : I feel bad about this, but so should they

Another transcript. The thing that came across in the talk that doesn’t translate to cold text is that this whole experience was amusing. My audience got it. The identity of the company is known to a few people, and I want it to remain discreet. First, I’m not in this business to embarrass people. Second, I’d quite like the choice about where and how I make a living to remain at least partly in my hands. Nevertheless, a true story this good cannot be passed up. While my treatment looks fairly shoddy, I don’t suspect malice: merely the usual consequence of happenstance meeting poor governance.

Update: With my own permission, here is a recording from a phone on the stage. You might still need the transcript. I had exactly five minutes, which is why the pace is pretty quick.

Can you smell that? Must’ve left a bridge in the oven. My topic at last year’s session needed to be handled carefully, so I read from a script. Here we go again.

In April, I resigned from ROLI to pursue a more intentionally fragmented career. My reasons aside, ROLI enters this talk only because the day after I gave my notice, I received an email from California. That coincidence begins this story.

The weeks that followed included hours of transatlantic phone calls, some easy, some hard. I spoke to ‘audio-famous’ people. The kind who have Wikipedia pages that they didn’t write themselves. These people had achieved a lot in their careers, before taking a corporate vow of silence and retreating from the world.

In June, arrangements were made to fly me to San Francisco, on the promise that I was about to meet a team of great people. I asked about those people. That information (even how many of them I’d meet) was classified.

To paraphrase, however, my brain contained something they might want to borrow. What this might be was a mystery then and remains a mystery now. They told me next to nothing and — joy of joys — gave me nothing to sign.

At the first of what turned out to be eight interviews, the opening question I was asked was ‘Would you please leave the building and come back in 45 minutes?’ This set the tone. The original itinerary was filleted to a fine mist. I was passed from hand to hand and back again, parked periodically on the sofa in reception to improve my acquaintance with their armed security guards.

The project coming together in the building around me was clearly a vast skunkworks, and I tried to piece together the details. I asked questions, naturally: What united cinematic colorists and a recently-acquired spatial audio company? Why was a famously expensive headquarters, ten or so blocks away, missing them? Why the hell, with all this talent on tap, would you fly a man over 5,000 miles only to do so little with him? That information was classified.

One of my interviewers was a co-founder and former CTO of a company that you’ve definitely heard of. On LinkedIn, his job title is now just a dash. I’ve been interviewed by a person who makes a living as a pregnant pause.

He played Bad Cop. Treating my lack of big corporate experience as a calculated affront to his dignity, one of his opening questions is why my career at ROLI had run in reverse. That’s harsh but actually fair: I’d prepared more than one answer to that one on the plane over. Five years after playing a central role getting the first few products out of the door, and four job title changes later, nobody really knew what I was doing there anymore, least of all me.

I didn’t tell him, although it’s true, that I was really after his job as a punctuation character.

Underscoring even those interviews that went well was a familiar note of crisis. Behind the scenes, phenomenal rank was being pulled. I quietly added some marginal notes to the chapter plan of my ‘narcissism in tech’ book.

Pencil sketch: The well of technology is looking somewhat dry in recent years. For want of a better plan, everything’s now driven by engagement data. Big tech is researching the psychological equivalent of crack cocaine while waiting to see if its competitors have a better idea. Little VC-backed companies have a similar problem. With no dependable long-term trends, a long term plan cannot be made to stick, so they grab the first thing they can, just like everyone else. What I saw in California taught me that this pattern extends as high and as wide as the eye can see.

For my entertainment, one of the smarter corners of one of the world’s wealthiest companies was candidly panicking its arse off about the kind of capital it would find down the side of an old armchair. You don’t stay rich by not speculating, but a pet project like this is always bait for executives to meddle.

As my first day in California drew to a close, I got to hang out with ten of my would-be colleagues: this started well, but went on until somewhat past 4am London time. My peripheral vision had started to fade. They were all called Steve. I ached from the mouth up and was numb from the chin down.

Halfway through the second day, I was told the job title I was interviewing for. I learned that I was in the frame because I’d spoken well at a conference some years previously. Not for the first time, I was genuinely impressed that any company had the capacity to review notes going back so far, but was no less confused about the action they had taken as a consequence.

And then, when I got home, nothing. My wife fretted about the possible change of continent while I periodically poked at the anthill across the pond. Four weeks later, I was informed that Bad Cop had the casting vote and I was out of the running for a position I hadn’t even applied for. Not to worry, though: they were scratching around for other opportunities.

In truth, I think I was used as a rubber duck. My purpose there was as an object that they could talk to to rescue their project. A friend of mine, insisting from the start that I should dodge this bullet, showed me a page from a tech-leaks website. There I found a plausible explanation, in the public domain, about what they were doing. It accounted for all of the people I’d met.

One morning, nearly two months after my trip, I awoke to a weirdly terse email from the person who’d first made contact: ‘I’m sorry but our plans do not include you’.

But by then, of course, the feeling was fucking mutual.

 

A tour of MPE, with apologies to Roger Linn

This article coincides with the publication of the MPE specification, which I facilitated and helped to write. If you register with The MIDI Association, you can download it free from here.

Roger Linn didn’t invent the drum machine, but he did make one of the first examples of a very good one. Drum machines began their lives in the 1960s, as little panels of buttons that gave organists a Latin rhythm to play against. The sounds of the day were built entirely out of snippets of simple waveforms and chuffs of filtered noise. These camp little drumbeats were subverted by Kraftwerk and post-punks, but hardly entered the cultural mainstream. Witness the Ace Tone Rhythm Ace.

When Donna Summer recorded ‘I Feel Love’ in 1977, the kick drum was provided by a (presumably both bored and anxious) real drummer; everything else on the percussion track was hand-crafted using a huge synthesiser. The result prefigured a coming of age for these early machines. Those who were present at the recording session routinely discuss what a pain it was to piece the single together, part by part. It was revolutionary in terms of its sound, and helped to cement George Moroder’s legacy as a producer, but the painstaking techniques he had developed to make such music possible would not catch on.

The LinnDrum, which arrived in 1982, actually sounded somewhat like a drum kit. Each drum could be tuned and tweaked like a recorded kit, to punch nicely through the mix. It, and contemporaries made by Roland (the company that Ace Tone became), Oberheim, and later others, underpinned the backing tracks of the synthpop music of the 1980s. These artificial instruments were intended to sound natural, but they were often programmed in a context that deliberately emphasised their inhuman nature. The machines began a long-lasting aesthetic of meticulously-layered, robotic backing tracks, where often the only live performer on the recording was the vocalist.

None of this would have been possible on such a scale without MIDI, a standard devised for interconnecting early synthesisers, which propelled this music into the mainstream by providing an easy way for drum machines, bass synthesisers, and other instruments to play their musical patterns in lockstep.

In person, Roger Linn has become fairly ambivalent about the Eighties, about the musical revolution he helped to create, and his obscurity as an inventor beyond the tiny music technology industry.

MIDI, too, has struggled with its reputation as a protocol for making robotic music. The early Nineties brought a thing called General MIDI, which enabled synthesisers to work more seamlessly with each other and computers so that, for example, orchestral music written to be replayed on one device would sound right on others. The unintended consequence of this was a craze for polyphonic ringtones about ten years later, when this technology became cheap and commoditised enough to end up in mobile phones. It has not been easy to foment much strategic consensus in the music tech industry beyond General MIDI, but that hasn’t stopped us trying.

Music tech has long had a problem where the tail wags the dog. The success of MIDI means that the form and function of electronic instruments has, for years, been driven by what the specification does best, and not the other way round. This is why, after a Cambrian explosion of exotic devices in the Eighties, all controller keyboards now look precisely the same: why there are always eight control knobs, and a data slider, and wheels marked ‘pitch’ and ‘mod’.

Why isn’t MIDI capable of supporting other approaches: of conveying and shaping natural human expression in all its variety? Well, it sort of is, but that’s not how it’s used. The limits of creativity are no longer determined by channel count, processing power, or data bandwidth. Rather embarrassingly, they’re driven by the cultural assumptions of a technical specification that’s more than 35 years old, and so entrenched in its own status quo that it cannot easily be changed.

This is why ROLI ran into difficulty about six years ago, when we tried to connect our prototype instruments to existing synthesisers, and quickly realised that not many people had given the problem of musical expression much thought. If every note in your performance needs to respond to a different movement of a user’s finger, there are technical workarounds that can achieve this. But, as hardly any instruments could generate such rich data, hardly any synthesisers could be made to work with it.

Other companies who had made expressive instruments had written their own synthesiser software, and ROLI got the situation under control by following the same path, and making Equator. We realised early on that customer confidence equals sales: that, only by building strong relationships with other companies, and making an ecosystem where our stuff would work, would we prosper.

A few synthesisers made by other people were actually ready to do the unusual things that we had to do, such as pitch-bending and swelling individual notes in a chord. After a couple of false starts trying to reinvent MIDI to fit our problem, we reinvented our problem to fit MIDI. In 2015, we started working with the MIDI Manufacturers Association to put together a specification called MPE [MIDI Polyphonic Expression]. This is a common way of allowing expressive controllers and synthesisers to communicate.

Aside from us, this has helped other hardware manufacturers to sell their instruments. Keith McMillen, Haken, Madrona, Eigenlabs, and Roger Linn benefit from MPE. It’s also helped makers of innovative synthesisers, such as Audio Modeling, find a small but interested market for their novel instruments. And it’s been supported by companies as far-ranging as Bitwig and Apple. But driving consensus among around twenty different vested interests has taken its time, and the specification has finally been published today, some three years later. You can find it and download it freely from the link at the top of the page.

If you already make electronic music, you probably won’t notice MPE compatibility slipping into the newer devices you use. This was one of the intentions. The other is that, if you do decide to go out and buy a Seaboard (if you haven’t already), there are a whole host of companies that have embraced MPE, and the Seaboard will work with their equipment.

Perhaps in an effort to produce a more interesting legacy, Roger Linn produces his own expressive instrument, the aptly-titled LinnStrument. He has been a champion of MPE since its pre-MIDI days. In his quest to imbue electronic music with the humanity that he unintentionally helped it to divest thirty years ago, he is making liberal use of the slogan ‘Stop using on-off switches!’

It’s my pleasure to say that I’ve stood on the shoulders of Roger Linn and others, in chairing the MPE Working Group, in writing and endlessly revising about 80% of the specification, and in speaking about it in various places. In glorious anonymity, it has been published today.

Now that you can afford to do so, and there’s a widely-supported and freely-published way of doing it, it’s become easier than ever to take Roger’s advice.