Thank You, Next: How Machine Learning Is Revolutionising The Way We Make Music

Luke Frost our consultant managing the role
Author: Luke Frost
Posting date: 2/7/2019 9:15 AM
From Vinyl to Tidal; we all know that the way we consume music has changed. Technological advances have made Steve Job’s claim that he would put “1,000 songs in our pockets” seem antiquated, whilst Spotify’s algorithms serve us tracks that we’ll love before we’ve discovered them ourselves. 

But can the technologies that have brought us these advancements change the way we make music? Whether it’s leading to new instruments or creating a song without our input, Artificial Intelligence is a game changer. 

Make Some Noise


Until recently, the best way to imitate a sound was by experimenting with the different settings on a keyboard. However, this is no longer the case, thanks to Google’s research arm Magenta. They’ve created the NSynth Super, an instrument that generates sounds based upon Deep Neural Network techniques. 

These algorithms allow the NSynth to not only imitate a sound, but consistently learn more and more about the specificities of that pitch, creating something closer to reality. Users can then combine those individual sounds to create something unique and entirely original. This is potentially just the beginning of a new wave of music, and in a decade’s time the NSynth could end up having as big an impact as autotune. 

Talking About AI Generation


Whilst we’re still waiting to see the impact of instruments akin to the NSynth, machine-led compositions are becoming more and more commonplace. Using a Recurrent Neural Network (RNN), one can feed a model existing music and ask it to generate something new. By learning the patterns and rhythms of notes from a variety of compositions, the model should be able to output an original and melodical sequence.

Although these may not be the most amazing tracks in the world, they do serve a purpose. Music production platform Jukedeck allows users to input their requirements for a piece of music (genre, temp, mood, length, instruments etc.) that can then be automatically generated using AI. Obviously these aren’t designed to be chart hits, but production music that can be purchased cost-efficiently for YouTubers, Short Films and other backing-tracks.  

Despite the fact that this remains the most common use of AI in music, some artists are looking to push this even further. Musician Taryn Southern, for example, has created an EP based purely on AI compositions generated using Amper Score. The platform generated a beat, melody and basic structure before Southern then rearranged and added lyrics too. Could this form of collaboration become the future of mainstream music?

Rage Against the Machine Learning


As with any change, AI’s interruption of the music industry is not without controversy, and there are those who believe that the human contribution is what makes music what it is. 

Indeed, there are still several limitations to what AI can achieve creatively. Despite a neural network’s success with creating original compositions, another’s ability to write lyrics was somewhat lacklustre. Despite being trained on a combination of lyrics (for structure), and literature (for vocabulary), its output was largely nonsense and included lines such as “I got monk that wear you good”.  

Perhaps, like Southern’s compositions, AI is best used as an accompanying tool. London-based start-up AI Music offer technology that ‘shape-shifts’ songs to adapt to the context in which they’re played. This could be anything from tempo changes to match a listener’s speed to remastering tracks to appeal to different moods and situations. IBM’s Watson Beat, on the other hand, creates compositions that naturally fit to the visuals of a video. In this context, as within many other industries, AI looks set to support our existing skillsets rather than replace jobs. 

Whether you’re looking to create collaborative technologies or revolutionise an industry, we may have a role for you. Take a look at our latest opportunities or get in touch with one of our specialist consultants to find out more.

Related blog & news

With over 10 years experience working solely in the Data & Analytics sector our consultants are able to offer detailed insights into the industry.

Visit our Blogs & News portal or check out the related posts below.

The Harnham 2019 Data & Analytics Salary Guide Is Here

We are thrilled to announce the launch of our 2019 UK, US and European Salary Guides. With over 3,000 respondents globally, this year’s guides are our largest and most insightful yet.  Looking at your responses, it is overwhelmingly clear that the Data & Analytics industry is continuing to thrive. This has led to an incredibly active market with 77% of respondents in the UK and Europe, and 72% in the US, willing to leave their role for the right opportunity.  Salary expectations remain high, although we’re seeing that candidates often expect 2-10% more than they actually achieve when moving between roles.  Globally, we’ve also seen a change in the reasons people give for leaving a position, with a lack of career progression overtaking an uncompetitive salary as the main reason for seeking a change.   There also remains plenty of room for industry improvement when looking at gender parity; the UK market is only 25% female and this falls to 23% in the US and 21% across the rest of Europe.  In addition to our findings, the guides also include insights into a variety of markets and recommendations for both those hiring, and those seeking a new role.  You can download your copies of the UK, US and European guides here.

Machine Learning: How AI Learns

Machine Learning: How AI Learns

Amazon has begun curating summer reading lists. How? Patterns. Facebook shows you ads for items you may have been searching for online. How? It learns from your browsing habits. Ever wondered how Facebook knows you were just looking at that pair of shoes or that particular guitar. The Data you feed it, feeds its brain. In other words, this is how Artificial Intelligence learns. Machine Learning. Whilst it can be disconcerting to know that a machine understands our buying habits, that’s not the only thing it’s used for. It’s also a pivotal tool in such areas as Bionformatics, Biostatistics, Computational Biology, Robotics, and more.  What is Machine Learning? Ultimately, it’s a method of Data Analysis which helps to automate model building and is part of Artificial Intelligence. In other words, it helps to solve Computational Biology problems by learning from Data to identify patterns and make decisions with little human intervention. This helps scientific researchers learn about many aspects of biology. However, running a Machine Learning project can be difficult for beginners, who may experience issues when trying to navigate the information without making mistakes or second guessing themselves. This is one of the reasons a Computational Biologist might want to upskill with a course or two in Machine Learning for a more robust understanding of the information being learned and applied.  The Good News and the Bad With each shift of industrial revolution, there has been one system which has made an indelible mark on our daily lives and the Fourth Industrial Revolution is no different. Just like we can no longer imagine factories without assembly lines, we can also no longer imagine not having Siri, Google Maps, or online recommendations. But, as exciting and as important as these things are, Machine Learning has become so crucial to our daily lives, so complex, it takes a technology expert to master it leaving it nearly inaccessible to those who could benefit from it. Why is Machine Learning Important? By building models to peel back the layers and discover connections, organisations can more easily and more quickly make improved decisions with little to no human intervention. Computational processing is both more affordable and more powerful. It’s possible to quickly scale and produce models which can analyse bigger and more complex data and there’s also a chance to identify opportunities and to help avoid any unknowns such as risk. Machine Learning is used in every industry from Retail to Financial Services to Healthcare. Here are just a few ways it has already transformed our world. Retail – Retailers are able to learn from their customers buying habits, predictive buying habits, how to personalise a shopping experience, price optimisation, and customer insights.Financial services – Machine Learning helps to prevent fraud and identify Data insights.Healthcare – Wearable devices allow for real-time data to assess a patient’s health. Medical professionals can also more quickly find red flags which can help improve diagnoses and treatment.Oil and gas – It cannot only help find where oil might be, but also predict refinery sensory failure, and streamline distribution.Transportation – Help to make routes more efficient and predict problems that could affect the bottom line. While humans can create at least one or two models a week; Machine Learning can create thousands.  Ultimately, the goal of Machine Learning is to understand the structure of Data. As it learns to determine what Data is needed for its structure, it can be easily automated and sift through Data until a pattern is found. This is how machines learn. If you’re looking to take your next step in the field of Machine Learning, we may have a role for you. Take a look at our latest opportunities, or get in touch to see if we can help you take that next step.

RELATED Jobs

Salary

£65000 - £70000 per annum

Location

London

Description

Fantastic deep learning role here - very collaborative research environment. The team are looking for an expert in neural networks.

Salary

£60000 - £70000 per annum + benefits + bonus

Location

London

Description

Are you looking for a collaborative team that you can learn a wide range of approaches from or a data-intelligent team that you can develop from?

Salary

US$140000 - US$160000 per year + Medical, Unlimited PTO

Location

New York

Description

Partnered with an exciting Fin-Tech company in New York City that build innovative AI solutions, looking for deep learning engineers.

Salary

500000kr - 600000kr per annum

Location

Copenhagen, Copenhagen Municipality

Description

The company is producing disruptive technology and is seeking to further advance their analytics function by adding a talented Machine Learning Engineer.

recently viewed jobs