Artificial intelligence is taking over our lives-and, it seems to be our music production. Let's insert the Matrix and learn more...

As the name suggests, machine learning is a form of AI that analyzes and stores data over time and then uses that data to make decisions and predict future outcomes. Deep learning is the next step in evolution: algorithms can automatically use a "neural network" similar to the human brain, without the need for human "supervision." In short, computer code lines can now be programmed to some degree for their own learning, and then use these learning to perform complex operations far beyond human capabilities.  

Considered to be the biggest advancement in software development in the past few years, this technology has benefited from revolutionary advancements in computing power and data storage, and has now become an indispensable part of daily life, just like Siri or Alexa smart storage. Data for predicting future actions. Ever wondered why Facebook’s "people you might know" and those nasty advice ads on social media are always so accurate? Weird, right? Before we even mention facial recognition software, spam filtering, image classification, fraud detection... 

Yes, machine learning algorithms are everywhere, and the music world is no exception. For our everyday music listeners in 2019, the streaming service algorithm drives a list of these suggestions to help you search for new songs and artists that you have never found. Last year, Google's Magenta research department developed the open source NSynth Super, a synthesizer driven by the NSynth algorithm that aims to create new sounds by learning the quality of existing sound.

At the same time, since Brian Eno's Koan-driven Generative Music 1 was released on floppy disks in 1990, computer-aided synthesis has always existed. Amper Music brings this concept into the 21st century: it is an automatic composition using deep learning Computer service-Generate music for a piece of media according to the user's choice of "style" or "emotion". Content creator Taryn Southern created a complete track based on Amper Music's artificial intelligence, and has since accumulated nearly 200 million scripts on YouTube.

In addition, this technology is being used to help music producers and performers. Audionamix's Xtrax Stems 2 uses cloud-based machine learning aids to deconstruct a fully blended stereo track into three sub-stems (vocals, drums, and music) that can then be used for live mixing and DJ mashups.

Whether you think these machines will take over our studios, it's clear that artificial intelligence technology still exists, and we are witnessing the beginning of the music technology revolution.

Does the robot have a musical dream? 

Software manufacturer iZotope uses machine learning technology in new versions of Ozone, Neutron and Nectar. Its chief technology officer Jonathan Bailey makes us feel...

These days, the terms "machine learning" and "deep learning" are widely used. What do they mean, in the words of laymen?

"Machine learning refers to a specific technology in the broader field of AI, which allows the system to find patterns in large amounts of data or make decisions based on previously invisible data. A common example is facial recognition technology. The software on mobile phones is clearly Never seen your photos-because they didn't exist before you took them-but it can recognize ('categorize') faces and groups ('cluster') them.

"Machine learning technology has existed for decades, mainly in the use of neural networks. Neural networks are connected statistical models that are inspired by the way neurons in the brain act as a system of connected nodes.  

“In the past decade, the combination of two forces has allowed breakthroughs in the use of machine learning technology: the explosive growth of digital data and the cheap availability of computing resources (due to cloud computing solutions such as Amazon Web Services). The use of learning. Deep learning refers to the use of highly complex neural network models that use multiple layers of nodes to connect in complex configurations, requiring powerful computers to train large data sets and operate.

Photography should kill painting. It does not. I believe in our ability to invent new ideas.

How does machine/deep learning help improve software tools for musicians and audio professionals? 

 "In the past few years, iZotope has invested heavily in these technologies. An example from Neutron, our smart channel strip, uses deep learning to identify ('categorize') which instrument the audio in any given track in a music session represents, and based on that category, and some of the things we analyze in it For additional acoustic characteristics audio, we recommend which dynamics, equalizer and/or exciter settings to apply to prepare the soundtrack for your mix.

“We are now using deep learning, not only to analyze audio content, but also to process it. In our recently released RX 7, music rebalance uses deep learning to “mix” the music mix into a single stem that can be rebalanced or Processing separately. We are exploring how to use deep learning to synthesize content in the future."

What are the big pros and cons?

“Deep learning solves some of the problems we have struggled with in the past. For example, many of our customers have asked us to use a method to eliminate the rustling of the lavalier microphones in the records, even with our powerful spectral analysis and processing techniques. . 

“It’s getting easier to use deep learning technology for companies interested in developing this feature, but it’s still not that easy. One of the main challenges in implementing a deep learning solution is getting good training data. Traditionally For companies that focus on algorithm development, this is a new area. Software for creating neural networks is provided free of charge, and commodity technology (Google TensorFlow is a common example). As I said, for a certain scale. For the company, getting a lot of computing power is reasonable. Data has become a huge bottleneck and raises an interesting question. Google offers their software and charges pennies for their cloud computing services, but they protect them closely data.   

"That said, deep learning is not a panacea. We still rely heavily on knowledge from digital signal processing specifications. Learning how to use and train deep neural networks effectively is becoming easier, but cutting-edge research is still done by highly skilled scientists ( Usually a Ph.D.). Neural networks can be difficult to debug, and sometimes they act as a kind of "black box"-you don’t fully know what’s going on inside. They are also computationally and resource-intensive, so make them Working in some real-time applications (such as synthesizers or audio plug-ins) is very challenging.

 “Deep learning is an exciting story, but in the end, we want our customers to get results that are not magic, but how she gets there.”

How do musicians use this technology while preserving creativity?

"There are several different research camps. A world of music, focusing on algorithmic music works. In this area, you own Amper Music, and their products can create music samples for your content, such as your YouTube videos or ads. Others focus on applications such as automatic accompaniment. As a result, some groups are trying to automate creativity while others are trying to strengthen creativity.

"This is a very delicate balance, but iZotope is firmly in the camp of enhancing creativity. I admire a research team like Google Magenta. Their goal is to use machine learning to create art-but this is not iZotope's philosophy or strategy. We I hope to use deep learning to help you create art. We are currently more focused on technology applications, but I do see us enter more creative fields as long as we stick to our creative goals. We are not trying to replace human creativity."

So will the software eventually write and mix our music for us?

"In some cases it is already. If you are a great creative singer, but you have never pioneered DAW in your life, deep learning will help you get a recording that sounds great without knowing what the compressor is. If you work all day at DAW, it will understand the effects you like and dislike, the visual and audible information you need to get the job done, and let you focus on the music itself.   

"Photography should kill painting. It doesn't. I believe in our ability to invent new ideas."

This article is from musicradar,Original address