Earlier this week, AWS launched DeepComposer, a set of web-based instruments for learning about AI to make music and a $99 MIDI keyboard for inputting melodies. That launch created a good bit of confusion, although, so we sat down with Mike Miller, the director of AWS’s AI Units group, to speak about the place DeepComposer suits into the corporate’s lineup of AI gadgets, which incorporates the DeepLens digital camera and the DeepRacer AI automotive, each of that are meant to teach builders about particular AI ideas, too.
The very first thing that’s vital to bear in mind right here is that DeepComposer is a learning instrument. It’s not meant for musicians — it’s meant for engineers who need to study generative AI. However AWS didn’t assist itself by calling this “the worlds first machine learning-enabled musical keyboard for developers.” The keyboard itself, in spite of everything, is simply a customary, fundamental MIDI keyboard. There’s no intelligence in it. All the AI work is occurring within the cloud.
“The goal here is to teach generative AI as one of the most interesting trends in machine learning in the last 10 years,” Miller instructed us. “We specifically told GANs, generative adversarial networks, where there are two networks that are trained together. The reason that’s interesting from our perspective for developers is that it’s very complicated and a lot of the things that developers learn about training machine learning models get jumbled up when you’re training two together.”
With DeepComposer, the developer steps by means of a strategy of learning the fundamentals. With the keyboard, you may enter a fundamental melody — however in the event you don’t have it, you can also use an on-screen keyboard to get began or use a few default melodies (assume Ode to Pleasure). From a sensible perspective, the system then goes out and generates a background observe for that melody based mostly on a musical fashion you select. To maintain issues easy, the system ignores some values from the keyboard, although, together with velocity (simply in case you wanted extra proof that this is not a keyboard for musicians). However extra importantly, builders can then additionally dig into the precise fashions the system generated — and even export them to a Jupyter pocket book.
For the aim of DeepComposer, the MIDI knowledge is simply one other knowledge supply to teach builders about GANs and SageMaker, AWS’s machine learning platform that powers DeepComposer behind the scenes.
“The advantage of using MIDI files and basing out training on MIDI is that the representation of the data that goes into the training is in a format that is actually the same representation of data in an image, for example,” defined Miller. “And so it’s actually very applicable and analogous, so as a developer look at that SageMaker notebook and understands the data formatting and how we pass the data in, that’s applicable to other domains as well.”
That’s why the instruments expose all the uncooked knowledge, too, together with loss features, analytics and the outcomes of the assorted fashions as they fight to get to an appropriate end result, and so forth. As a result of this is clearly a instrument for producing music, it’ll additionally expose among the knowledge in regards to the music, like pitch and empty bars.
“We believe that as developers get into the SageMaker models, they’ll see that, hey, I can apply this to other domains and I can take this and make it my own and see what I can generate,” stated Miller.
Having heard the outcomes thus far, I feel it’s protected to say that DeepComposer gained’t produce any hits quickly. It appears fairly good at creating a drum observe, however bass strains appear a bit erratic. Nonetheless, it’s a cool demo of this machine learning method, regardless that my guess is that its success will likely be a bit extra restricted than DeepRacer, which is a idea that is a bit simpler to perceive for many because the majority of builders will have a look at it, assume they want to have the opportunity to play an instrument to use it, and transfer on.
Further reporting by Ron Miller.