• Creation

Finding the Heart and Soul in Artificial Intelligence: a journey to create music with AI

  • Creation

Finding the Heart and Soul in Artificial Intelligence: a journey to create music with AI

The concept of artificial intelligence (AI) has fascinated humans for centuries, from the intelligent robots mythologized in ancient Greek, Chinese, Egyptian and Arabic societies, to modern pop culture’s futuristic obsession.

We’re equal parts intrigued and terrified of the unknown implications of the advancement of intelligent machines. Fritz Lang’s 1927 groundbreaking film, Metropolis, delivered a chilling vision of 2026 where a human-like robot is used to eliminate the working class, while the 1999 film, The Matrix, depicts a frightening future of human enslavement to villainous, superintelligent AI programs.

We’ve watched HAL, the sentient computer (2001: A Space Odyssey, 1968), Sonny,  the anthropomorphic servant droid (I, Robot, 2004), and more recently Ava, the beautiful and deceptive robot (Ex Machina, 2014), challenge what it means to be human, as the presence of artificial intelligence continues to increase in our everyday lives.

Fortunately, we’re not yet under attack by an army of self-aware cyborgs (experts fiercely debate the arrival of ‘superintelligent machines’ exceeding the cognitive ability of humans within the next 100 years), but you’re likely using AI everyday to search the web, shop online, talk to friends, and get around town.

Put simply, artificial intelligence is ‘intelligent systems that have been taught or learned how to carry out specific tasks without being explicitly programmed how to do so’. Everyday examples of AI include smart assistants like Siri and Alexa, Google’s AI-powered maps, facial recognition to unlock your phone, ridesharing apps like Uber and Lyft, and spam filters in your email.

AI isn’t always the answer.

AI is responsible for some amazing advances in healthcare, including robotic surgery, medical imaging and disease prediction, and it’s expected to make some incredible developments in 2018, including the arrival of driverless cars. Even commercial airlines are adopting AI autopilots – there is now only seven minutes of human-steered flight on the average Boeing plane flight.

So when I decided to build a music composition tool using AI, I was shocked to find that AI isn’t always the answer. It sounds counterintuitive, however AI’s limited development can present a hurdle for the tool you are trying to create. The use of AI can result in both good or bad innovations, but how can you predict this in advance before spending a ton of money to create something that people don’t want?

What is a good innovation?

I began to answer this question by delving into the parameters of a good innovation and comparing that to my experience building Juxtapus, an AI driven music composition app.

Innovation is about improvement and and achieving change by creating better processes and products. Innovation relies on creative and original thinking however most innovations, whether they are good or bad, aren’t completely new; they are a substitute for something that already existed.

For example, we can track the evolution of human storytelling through our innovations, from oral mediums to visual, written and printed substitutes. The earliest known forms of storytelling can be found in cave paintings across the globe, dating back to 30,000 years. Prehistoric artists used a range of earth pigments, such as charcoal and minerals, to create vivid paintings believed to be used for sharing information and ceremonial purposes.

Cave paintings were replaced by various innovations throughout the centuries including the quill, the most effective writing tool from 600 AD to the 19th century. While the ballpoint pen may have been a groundbreaking tool when it was patented in 1888, it was simply a better substitute for the quill.

Today, we use storytelling mediums such as film and television as our modern substitutes for those early cave paintings. Instead of huddling around a fire as a story is woven with paint, we now gather around the lounge and listen to stories through our screens.

To determine the worth of a substitute – and whether the innovation is good – you have to evaluate it’s improvement on the previous tool. Television and film are far more effective at telling stories than early cave paintings, therefore they’re a great innovation.

This is the standard to which I hold myself and my company as we continue to solve the problem between AI and music composition, with the aim of helping people to make music.

What are we substituting?

The best previous alternative to our AI driven music composition app, was to formally study music composition, spending many years learning the rules and art of composition. You would attend a music school, take private lessons, or specialise in a certain genre, developing a brilliant skill with many advantages.

Not only would you know how to compose a song, but you’d have the ability to express any ideas that popped into your head. You could listen to a movie soundtrack, understand how it was created, and borrow and manipulate ideas from that music.

Your only limit is your creativity.

You would possess autonomy, freedom, control and flexibility. You could convey the abstract and express yourself, dreaming up blissful harmonies, mournful melodies or whimsical symphonies. You could depict your day with a chord, or relay a feeling with a verse. Your only limit is your creativity, courage, desire and inspiration.

We are aiming to substitute the skills gained from the formal study of music composition with our app. If our innovation cannot bestow all of these abilities on the artist, then it will not be a worthy substitute.

The problem with AI

With this overarching goal in mind, I sat down to talk with AI experts about what we could build. With every meeting, I noticed a pattern of themes in the conversation. The focus drifted from ‘How can we bestow these features and abilities to the artist?’ to ‘How can we create a user experience that gets the user to give the AI the information it needs to perform?’

The biggest limitation of AI is that it’s only as smart as the given data. Unlike humans, AI cannot yet integrate knowledge and it requires the user to provide a ton of information so it can learn and perform. Ultimately, the user experience is about bending to cater to the limitations of AI.

I realised the focus was not on meeting the needs of the human using the software, but on meeting the needs of limited AI with learning difficulties. The driving goal was to force humans to behave in a particular way so the AI could learn and perform.

But there was a problem – we were ignoring the element of human desire and need. If we had continued to dismiss this problem, instead of achieving the goal of creating a substitute for being a composer, we would have created a substitute for hiring someone to compose music on your behalf.

With such a system, you would tell the AI to make something new or change a beat, and once you’re happy with the completed song, you would take the music and pretend you composed it yourself. This is a common trend in the music industry, and many of the biggest hits today are devised by a team of background composers, rather than the actual singer.

Manufactured music vs creative freedom

Using manufactured music or hiring a ghost composer may suit some, but the truth is, people like having creative freedom. Whether you’re a multi-million dollar celebrity topping the charts, or a novice musician learning the ropes, people like to be in control with the autonomy to express themselves.

The best music flows from the heart and soul, so it’s likely you don’t want someone else representing or interpreting the very core of your being. You’d rather be in the driver’s seat, creating genuine music with your great ideas. I realised that the first step to building an AI driven music composition app was not to build the AI, but to ensure we strengthened the capacity for humans to be creative.

We want to allow you to express exactly what you want to say, using the AI to help you communicate musically. You’ll begin a dialogue with the AI in an effective way, so the AI will understand what you want to create.

If the AI doesn’t deliver what you intended, you can intervene, and unlike an ancient Roman passing judgement on a defeated gladiator, you won’t simply put a thumbs up or thumbs down. You’ll be entwined in the process, with primary influence over the music.

You could spend a few seconds using AI to create a manufactured song that you don’t really care about, or you can spend minutes or hours crafting a piece of music that flows from your heart, proudly conceived with your real ideas and influence.

The app will give you control over the music you make but more than that, we also want you to feel empowered.

The power to control AI

There’s a difference between power and control.

Imagine you’re sitting in a car with your hands on the steering wheel and you’re in control of the vehicle. But you don’t know how to start the ignition or change gears, so you’re stuck in the driveway stalling the car. You may be in control of the car, but without the power to start action, you cannot go anywhere.

Power is the ability to do something in a particular way and the same concept applies to AI – you can control the system by providing data, but if you can’t influence the AI to behave the way you want, you don’t have any power.

You can control the system.

So our first step is to give you the ability to compose and strengthen your capacity to be creative; the next step is to ensure that your composition is benefiting from the best musical ideas out there.

For example, the AI will have the ability to identify the hottest trends and styles and you’ll  have the option to call on the AI to find a certain drum beat or guitar riff to complement your song, while also using the composition tools to express your original musical ideas.

With the power to wield the music in your mind and the choice to apply the musical genius of others, you’ll have everything you need to create your perfect song.

The future of AI

Artificial intelligence is often synonymous with a sinister future, largely due to Hollywood’s grim portrayals of robot uprisings and human extermination, and also ominous warnings from the world’s greatest minds.

Elon Musk, creator of Tesla and SpaceX, Bill Gates, founder of Microsoft, and renowned physicist Stephen Hawking, have all voiced concerns about the rise of super artificial intelligence, with Musk equating the improvement of artificial intelligence with “summoning the demon”, calling it the biggest existential threat to the human race at an MIT conference.

However, there’s a strong beneficial-AI movement working to overturn alarmist AI myths and research AI safety, ensuring AI remains beneficial by amplifying human intelligence and helping human civilization to thrive. Intelligence equals control, and if humans align the goals of superintelligent AI with ours, and remain the smartest beings on earth, it is likely we will always retain power over our superintelligent creations.

For now, we don’t have to worry about sentient robots turning evil. Rather, we should encourage and accelerate the incredible AI advancements of humankind, alongside the development of AI safety strategies.

As we’ve developed our AI driven music composition app, we’ve recognised both the limitations of AI and it’s benefit to our ability to create incredible music.

We’ve also discovered something crucial; in order to get the best out of AI we must focus on the element of human desire and need. The goal of building AI must come second to the top priority of empowering human ability.

If we remember the humans behind the machine, we’ll find the heart and soul of artificial intelligence.

Leave a comment

Your email address will not be published. Required fields are marked *

Leave a comment

Your email address will not be published. Required fields are marked *

Sign up to Juxtapus for a 3-month free trial and start making music like a pro.

Try it now