musicube blog image

AI for dummies

February 15th, 2022

For decades, artificial intelligence has been raised as a topic in science fiction books and in social discussion. When we listen to the term artificial intelligence, there are a lot of ideas that might come to our mind. For some people it will directly link to smart robots, feeling emotions and aiming to become humans. For other people, it might link to privacy issues, intelligence agencies filtering out keywords from our digital conversations. For others, it will link to the search engines, or the smartphones being able to recognize faces or suggesting us to purchase whatever product we might have been thinking about recently. It might be also Spotify, YouTube or Netflix, suggesting content that could potentially interest us, and a long list of etcetera’s. But, how much of that is artificial intelligence? And how much it is not?

In this article we would like to bring up a little bit of historical context to the artificial intelligence and its evolution along time. We will also talk about the definition of artificial intelligence, and why it has always been a controversial concept. Finally, we would like to focus on how AI is being used in creative industries, especially in music. We will also talk about how does musicube’s AI engine work, and how it can help us to find the music we are looking for. By the end of this read, we hope everybody has learnt something new and interesting about AI as a general concept, and about the cool things our AI can do for you here in musicube.

1) History of AI

AI sounds modern and disruptive. It sounds to Science Fiction and inspire a future, modern world. But the term is way older than you might imagine. The first person who considered it (without yest inventing the term), was Alan Turing in the 1950s, when the essay “Computing Machinery and Intelligence” was published, and he asked the question “Can machines think?”. It was not until 6 years later, where the term artificial intelligence was first presented by John McCarthy at the Dartmouth College. The coining of the term lead to deep considerations about the definition and applications or artificial intelligence and triggered the studies on the field. Three years later, in 1959, the first “Artificial Intelligence” laboratory started to run on the Massachusetts Institute of Technology (MIT), in the USA.

Since AI is tightly linked to the technological advancements and the capacity to process great amounts of information, its evolution was not smooth. It wasn´t until the boom of the internet when experts consider the first real wave of AI development started. The second wave of AI evolution occurred with the development of what it is considered “mobile” internet. In this period, a change of paradigm occurred, and we shifted from an internet based on “static” information (normally outdated) to an internet working with real-time data. This altered profoundly the way of conceiving AI systems, and the way they operated.

The third wave is linked to the development of the Internet of Things, where a huge number of users and devices, equipped with different types of sensors, are starting to interact with the real world and take decisions based in the observation of these variables. The fourth wave is considered to be on its way, and it is expected to be the one in which true artificial intelligence will start to emerge. This brings up a question: we’ve been talking about the history of AI, its evolution and waves, but: what is indeed the artificial intelligence? How do we define it

2) What’s the AI

The truth is that the concept of AI is really hard to define. It is commonly considered “the ability of computer systems (both hardware and software), to perform tasks that normally require human beings to use their intelligence”. Nonetheless, a lot of experts can’t fully agree on this one. The reason of these disagreements is not the “artificial” component of the word, but the “intelligence” part of it. This happens because it does not exist a clear definition about what it is, in fact, the human intelligence.

When the concept of “intelligence” is used in connection to human beings, it normally refers to skills such as learning, planning, reasoning, making decisions, solving problems, perceiving, and conceptualizing the world, and expressing that experience in form of language and other creative ways of expression. The concepts of mind and consciousness are also related to human intelligence. Thus, artificial intelligence is a very broad term, that could be linked to all the functions that were mentioned in the first section of this article.

There’s so much to develop and clarify in the field of artificial intelligence. Although the development of artificial intelligence is still a work in progress, it is being already applied in different fields. One of them, and the one we feel more motivated about (because it is what we do!), is in entertainment and creative industries.

3) AI in entertainment and creative industries

Artificial intelligence has a big impact in creative industries. It is being used for several tasks. There are a lot of entertainment platforms (Spotify, Netflix, YouTube…) that use AI to suggest matches based on the users’ personal preferences and on other users’ past choices. Other AI engines as specialized on analyzing the audiovisual information to identify plagiarism cases, or, simply, to be able to tell the user which song is sounding in the background, or which song are they singing or humming. In the music production world, there are also AI engines specialized in assisting songwriters in the composition or mixing duties. Each of these functions imply very sophisticated and specialized design and training of the AI to be able to succeed on the task.

4) AI in musicube

In musicube, we are using AI at different levels to perform functions such as suggesting new tracks inspired on a given one, retrieving and curating metadata information from different sources, and analyzing sound characteristics of the tracks to tag them properly and facilitate searches within a song catalogue. All these functions are aimed to assist music professionals and other music lovers in discovering the track or tracks they are looking for in an easy, fast and effortless manner.

Our AI motor uses a mathematical transformation to reduce the complexity of a soundwave into a limited number of measurable and objective variables, such as tempo and key. Our algorithm also retrieves data from different resources. This allows to identify another huge amount of metadata not directly embedded in the track’s soundwave. Thanks to this “retrieval” function we can identify contributing artists of the track, the nature of the vocals and instruments (and dominant instruments) present on the mix, the mood, the music genre and many others.

After years of training and perfecting the engine, now, in musicube we can automatically detect, validate, and annotate many different musical features of the track in no time. Some of them, such as key, tempo and mood, are well known and understood by average public. But we can also identify others such as the arousal of the track, the engagement it has with the listener and its pleasantness. But this is not all. We can also identify harmony, scale, tonality, roughness, timbre, rhythm, space, groovyness and many other awesome variables that we’ll be talking about in more detail in an incoming post.

When it comes to add metadata to a track within a catalogue, the completer and more detailed you can be, the better for facilitating yourself and other users, to find “the track” that is needed in a given moment. Manually annotating rich metadata in a track is time consuming, and it requires a huge amount dedication and expertise. The only reason why we are able to do it for you in no time in musicube, is because we put together a great multidisciplinary team of computer engineers, data scientists, musicologists and musicians, working together with the common goal of developing a very sophisticated AI ready to serve its users in the best possible manner. And we are always working hard to improve.

Would you like to see for yourself how our AI tags your songs? Then register at musicu.be. We will be happy to show you.