Unicorns in Business Suits

There are staples in every engineer's plugin repo that are just lying there, looking up at them disparagingly from the bottom of their toolboxes: Melodyne, Izotope RX, Antares Auto-Tune; some of them are both beloved and hated, by purists who prefer to use only Tim Petherick Nebula 4 hardware emulations for the most authentically analog experience. Fine. I was one of them. But what could possibly benefit me from parting from tradition? Andy Wallace, heck, even Al Scmitt romanticize simplicity. Al tells us that the higher up one goes in the field of audio engineering, the easier the job gets. At a high profile studio with bigger budgets, that means better musicians, better instruments, better room acoustics, better microphones, preamps, cables, converters, drudgery being taken care of by technicians and unpaid interns so that the "engineer" can make his art while minimizing distractions.

It seems like we as engineers tend to romanticize our tools more than our art sometimes. I have a proposition to make: audio engineering is not "real" engineering but a name we give ourselves to fit into society, slightly higher than the lowly artist continuously begging for scraps across each generation. The audio engineer is a new phenomena that emerged alongside the "soft" sciences that protrude and blemish (not embellish) our universities.

My point is that we either need to stop taking ourselves so seriously (since we're just artists with more sophisticated tools)—or—we take our work as seriously as the most rigorous of scientists.

From here, I am led to an epistemological question that interests me greatly: what is the origin of creativity? I have no clue how to find that answer. Intuition makes me gravitate towards a far-reaching, multidisciplinary explanation such as harmonic analysis, as another question closer to my current capabilities arises: if we can construct anything that we understand (Maybe this isn't true, I'm inverting the Feynman quote "What I cannot create, I do not understand."), why haven't we figured out how to accurately replicate the transforms from the input to the output stages of an analog piece of hardware, everytime? Some exceptional plugins come to mind: Acustica Audio Magenta, TimP's STA-Level, Universal Audio's Pultec, Airwindows ToTape—but why don't these guys just keep batting a thousand? Hit after plugin hit? Different methods yield different results.

Some developers implement a "black box" approach that may use entirely different methods of getting to the end result than what was actually accomplished through hardware. Neural networks come to mind as one of these approaches. While being much more efficient than traditional white box C++ approaches, the "issue" I people may have is one of morality. To design a neural network is akin to designing a Prius, and now the only thing left to improve upon is efficiency. Correcting a loss function is not romantic. There's no blood involved, so where does the innovation lie?

While it may not be obvious, neural nets are actually complicated. In recreating transformations circuitry can affect, they can implement a similar structure of: filtering, nonlinear transformation, and recreation. This can recreate the desired nonlinear effect (e.g. distortion) by emulating the philosophy of the physics behind a piece of hardware. So it's a "black box" in that it's not recreating the physics in a series of logical steps like a circuit diagram would show, but if we can understand (or empathize with) the similarities in design, we can proceed with less doubt about the methods.