Name: Ravin Mehta
Occupation: Producer, label founder at Made By Robots
Current release: Made by Robots has built up a diverse catalogue of releases, including the Lofi Tunes series, or, more recently, "Ritmo Acido" by GPU and BERF.

If you enjoyed this Ravin Mehta interview and would like to know more about the label, visit the official website of Made by Robots. The label is also in Instagram.



How has the debate on AI and music evolved, do you feel?

We are currently at the point where AI got widely acknowledged for doing impressive things, be it with textual, visual or audio content. This has produced mixed reactions and for AI to become an accepted creative opportunity, the obvious and justified criticisms and concerns need to be addressed.

In terms of technological progress, what have been some of the points over the past decades where you personally had the feeling that something really had shifted?

Technological progress inspired and enabled me to start as a writer and producer back in the day. In the 1990s pro quality recording trickled down into the home studio with technologies like sampling, sequencing and drum computing and with gear like the Atari ST, Alesis ADAT, the S1000 or the 303. This enabled a new generation of music producers to shape the future sound and invent now-mainstream genres like house, techno or hip hop.

With modern laptops everything required to produce music from idea to master and even distribution can be done inside the infamous “box” without ever seeing the light of day - which I think is another major shift.

These are mainly the bad feeling of AI stealing from established artists (by using their data for training) and the lack of an agreed way to apply control and compensation.

What is the difference between the use of generative AI and a 303, for example?

Today, the new possibilities of generative AI in music seem endless and results are anywhere from boring to outlandish and crazy.

The difference with AI is that it’s not only happening inside the music industry (like sampling for example). AI music is embedded into a broader consumer AI trend and therefore does not have the luxury of growing in privacy and letting the results speak for themselves before focussing on legal and commercial aspects.

What's your take on the balance between technology and creativity?

First, creators and listeners still decide what they want to listen to and therefore curate what’s being called music. AI might contribute to or even come up with new musical ideas but these ideas still need to be interesting and attractive to humans.

For creators, the technological palette that can be used to make music is getting extended by AI tools and maybe you’ll end up with a Python programmer in your next band.

But I remember producers hiring Fairlight CMI or NED Synclavier programmers to create hit songs …

Since you've worked closely with labels for a long time, what is your impression of how they see the development?

My perception is that labels know and see the opportunity of monetizing on AI companies using their copyrighted catalogs of masters and songs. From a commercial perspective AI looks like on-top money.

But at the same time it seems that nobody wants to be the first to come up with something constructive. And since many prominent artists loudly advocate against AI, I guess nobody wants to alienate their existing artist roster.

Tell me a bit about how the idea for a label with a strong AI-conceptuality came up and what the most important considerations for Made by Robots were when setting it up?

After being in the music industry for many years after school, I moved on to the digital space when the internet started. Also, my previous company (The unbelievable Machine Company) was one of the first AI companies in Germany.

So I thought I could finally combine all of these things into a new concept that became Made By Robots. It’s a technology/music creation hybrid that runs Made By Robots as a label that is distributed by Sony Music’s The Orchard but also a song rights collection space (Published By Robots) and a software development thing (that has no funky name yet …).

The website refers to Made by Robots as a “next-generation record label.” In terms of the label work, what is different compared to previous generations?

Today in the independent space for electronic music, the typical record label specializes in releasing and marketing finished products.

Made By Robots not only provides that but also invites and helps artists who want to explore the AI space or need additional production skills to turn great ideas into great products. We’re especially successful with DJs who have ideas but do not have the time and or skills to turn these ideas into finished products.

On the marketing and promotion side we are actively seeking new ways to market tracks outside the traditional boundaries of the music industry - the more out-of-the-box the better. Also we leverage our technology knowhow to utilize things like APIs and data analytics to improve automation and optimize budgets.

But I can also spend days goofing around with new software toys from the recesses of the internet.

Which functionalities of AI tools have you personally found to be most promising and fruitful?

I make great use of some self-written software for co-writing new songs. There are some interesting tools for sound creation out there but I think there are already too many sounds and do not use them. AI Mastering is working pretty well but personally I still do it manually.

All-in services like Suno or Udio are super impressive tools but first need to sort out their ethical topics (as said above), before they can be used in a real production.

For marketing ChatGPT and other image and text generators help a lot to streamline output when you take the time to quality control and curate the results.

What can you say about what your proprietary tools do?

The openly available AI software platforms either face the challenges mentioned above or restrict the use of the results to R&D. So for now the only way to be safe for commercial use is to either create the data yourself or to license the raw data and not the trained model.

Essentially we have created a data pipeline that can take any kind of musical (or not so musical) data and train a model that can then be used to create something new. Examples are the tech house tracks on the label but also the music made from data we got from sensors on plants that we used to train an AI model (and then make music with it).

In terms of the legal aspects – what are currently the main legal challenges of making music with AI?

As said above, the main challenge that needs to be managed is the legal and procedural mechanics of compensation for the use of copyrighted material.

And while the bad feelings that are currently associated with AI music might not be a legal issue, they need to be addressed in the same process.

From your point of view, should pure AI compositions be copyrightable?

Here the answer lies in the previous questions.

If the whole chain of used copyright is contractually defined, then the output should be copyrightable as well. The more interesting question is probably by whom …

The term “ethically clean” is intriguing in relation to AI as the ethical implications of this technology have been on many artists' minds. What does this entail, specifically?

Overall to not infringe on anybody's rights either by obtaining their consent or by using self generated and/or synthetic data that is free of third party rights.

Also trustful acting and explaining the building blocks of AI to artists and co-producers as much as possible is necessary.

What is the pleasure and allure of working with AI tools for yourself and the artists you've been in touch with?

I meet a lot of people who are interested in AI but don’t know how to include whatever tool into their production. Also how to use AI in an ethical way. To co-create with these artists is a lot of fun and probably for now the most satisfying and productive thing to do.

One big question resulting from the use of "intelligent" production tools and AI is whether the results are as important (or even more important) as the process (and the joy of creating). How do you see that yourself?

My tech nerd side is just amazed by what’s already possible and makes me want to play around with it all day - just because I can.

While there might be use cases for automated music creation, maybe in advertising, in-store background music or in elevators, the fun part for me is to co-create something that I or we did not think about beforehand. And that might be very accessible or very experimental.

Is your experience that AI reduces human involvement in the creative process – or does it conversely increase it?

I think AI just changes human involvement. Drum computers enabled us to think in a more technical way about drums and required programming skills instead of hitting a real drum kit.

AI requires new skills like prompt engineering or programming for example but it just means that new people get involved in music creation.

From what you've heard so far, how is AI music different from that of humans – can it be?

Full tracks from the large AI music platforms are clearly not on par with human production but getting there. Since being trained on original data their models produce either mediocre approximations of existing things or need human involvement to turn a promising idea into something great.

Personally I think the most interesting things come out of the use of non-musical data for music generation.