Great links, Rob. I would be totally happy to surrender some of the less glamorous, incidental audio in a given edit (e.g. foliage brushing or debris in an action sequence) to an algorithm that’s learned to put one over on the majority of listeners if it buys me more time to make the cool stuff sound cooler. That’s not to say that there isn’t craft in all types of sound editing, or that you’d want to blanket hand over a type of sound in all contexts; in a delicate close-up, I would rather handle the cloth track myself.

But you can see that there’s already some enthusiasm for bringing up synthesized variation / sound creation of the type you’re mentioning in the game community with modal synthesis methods such as Wwise’s SoundSeed Impact, little-used as it still may be. Wouldn’t doubt that Rockstar’s RAGE engine is capable of lots of the same sort of thing. We’re pretty close to capable of feeding the parameters those synthesis models need to match object type and scale already, it just seems like the program you linked is able to do it from the ground up.

It’s tough to say where that process stops, though, and where we decide to draw the line on employing machine learning-driven sound synthesis / editing. Crazy to think that’s already a few clicks past what I was envisioning in this article, and it’s being written about today. Crazier still that all the funding which is driving it doesn’t seem to be coming from some of the entertainment industries it could come to affect the most. I’m guessing there’ll be lots of initial resistance to methods like this but a new generation will bring it into the workflow piece by piece.

Similarly, LANDR strikes me as a Ozone-esque solution which has access to mountains of data and can tune itself–but Ozone already came packed with presets that were / are ‘good enough’ for most folks without understanding. The introduction of both tools arguably raises the overall music production quality bar without having improved creativity across the field, but, still a net win. Perhaps the future of our education moves more strictly towards fostering creativity and all that which makes us human.

I think a good question to ask is — does machine learning solve a problem for sound designers? I recall this news release:
http://news.mit.edu/2016/artificial-intelligence-produces-realistic-sounds-0613#.V1-OxQWangg.facebook
Could this tool solve a design problem?

Perhaps there are technical solutions like what LANDR is doing with mastering: https://www.landr.com/en. The technical side of design is probably more likely where this innovation can be used, since it is less subjective. For example, if Pro Tools could learn the patterns in which I edit audio and eventually just trim and fade automatically or learn how I use other effects and plugins, this could perhaps speed up certain design processes.

But at the end of the day I’ll keep the machines at arm’s reach from my creative process.