Tech+Art Podcast: Monica Dinculescu

LISTEN HERE

You can catch Tech+Art Podcast in the following places – or your favorite podcast app:

OVERVIEW

Welcome to the new Tech+Art Podcast!

Join us on this adventure as we meet & speak with: artists, makers, researchers, designers and creators from all background and fields.

Our objective is to understand their creative perspective, dive into their workflow & creative process, be inspired by new ideas and their work – and stay one step ahead of cutting-edge industry developments.

OVERVIEW

Welcome to the new Tech+Art Podcast!

Join us on this adventure as we meet & speak with: artists, makers, researchers, designers and creators from all background and fields.

Our objective is to understand their creative perspective, dive into their workflow & creative process, be inspired by new ideas and their work – and stay one step ahead of cutting-edge industry developments.

If you have no knowledge of machine learning, if you don’t care how the algorithm works, you should still be able to use it.

IN THIS EPISODE...

In this episode, we’re chatting with Monica Dinculescu, an Emojineer currently at Google Brain, where she works on Magenta, an open source research project making music and art with Machine Learning.

Monica joins us to share her story, her approach to fun and creative “silly projects”, and how she approaches turning machine learning into simple tools to empower other creators.

You can follow Monica’s work online:

There’s a full list of all resources mentioned at the bottom of the post!

Question 1: Today you work at Google Brain on the Magenta.js project - can you tell us a little bit more about both?

All the research that we do is about enhancing people’s process with machine learning. People like artists or musicians or creators – we’re trying to apply machine learning to them in ways that are actually useful.

Art is good. We’ve been making art for thousands of years. We’ve been making music for thousands of years – but the tools that we use to do these things haven’t really progressed […] we haven’t fundamentally changed any of these things. And that’s what we’re looking at. Building smart instruments that can adapt to you, that can sound like you, that can help you to get out of a creative rut.

Question 2: What do you think is missing from the industry? What do you wish existed?

There’s a lot of things that are missing […] I’m a big cheerleader for the web, I think everything should be on the web. I think web is the easiest medium to transfer. Audio on the web is really hard – like genuinely really hard. Getting microphone input that works in all your browsers is really hard […] Because I think the market for building music apps on the web is really small right now, browsers don’t focus on that very much – and that’s kinda frustrating from the side of somebody who is trying to build music apps for the web.

With Magenta.js in particular, […] we try to pry all the machine learning away basically. The API is very straightforward. It gives you a notes sequence which is an arbitrary representation of music and you don’t have to worry about the math behind it […] You basically ask it to give you a sample and it gives you a sample.

If you have no knowledge of machine learning, if you don’t care how the algorithm works, you should still be able to use it. There’s a ton of engineers who are really creative and a lot of the times we expect them to understand all of this math to work with machine learning and I think that’s absurd. […] We try to take a lot of those pain points away (I hope).

Question 3: What are "silly ideas" and why are you so passionate about them?

I call them silly because if you don’t call them silly and you take them seriously, then people treat them very differently. We live in a society that’s taught us to judge things. So I’ve been doing art all of my life – a lot of the art that I look at I’m like ‘oh, could I have made that? That means it’s not really good’. Which is a terrible way to approach anything.

So I don’t call them any of that. I call them ‘silly’. Because when you say, ‘Look at this silly thing that I built, it makes music out of bananas’ or something like that… People go in and want to have fun – cause you’ve already told them it’s going to have fun. […] And that’s really what I want people to do. I want people to enjoy the thing that I’ve made ultimately. So as a result, I build all sorts of things that basically let you produce something. Portraits with emojis. Or music….

Signup for our newsletter to stay connected!

Your dose of creativity

Question 4: Where do those ideas come from? Is it a pain point that you are trying to solve?

It’s almost never a pain point. So apart from silliness, I’m really big on rules. So for a while (before I came on the Magenta team), basically all of the silly things that I built had produced emoji in some sort of way. That was basically my goal, I really wanted to work with emoji because I thought it was really interesting at the time. So as a result, I would make visual things for emoji. And all of the questions there would be like ‘what can I build if the target is emoji?’ and then the things that you have to focus on at that point are the experience that you’re building not like what it’s gonna look like – because you already know what it’s going to look like, it was full of emoji. And now because I work with music, my rules are that I pretty much only work with things that take in MIDI or generate MIDI or try to do something with MIDI.

One of my favorite things that I’ve ever built in my life is this project called MIDIcity 2000 – that basically takes MIDI files and shows them as a cityscape – with like different rows of building. Then you get to control those building by increasing the density between them or removing out some instruments and then that makes another melody out of it because you’re just manipulating this music representation.

And then I have like burst of time where I do nothing. There will be like 3 months and I’ll be like ‘I have no ideas’.

Question 5: What do you think the next step is in terms of helping all types of creators?

So I think that’s a hard question, mostly because we’re not a product team. So we’re a couple of engineers and a bunch of researchers and this is not really how you build products. This is why we open source all of the models, because I think there are better suited teams to do this. We don’t have designers. We don’t have product people. We have us. So I’m not sure that we are necessarily the people who should be solving all of these problems. I think we are the people who should be helping people solve these problems, but not necessarily building these apps that are teaching people how to play music […] That being said, I like to think that a lot of the demos that I build and that the team produces are very accessible to non-professional musicians especially. Because I’m not a professional musician, but I want to have fun with music.

Question 6: What is some advice you would share with others or a younger version of yourself?

I think the biggest thing that you learn as a senior engineer is how to find things – and sort of accepting that you don’t know the answers. When I was a junior, I was really intimidated by the fact that like, everybody knew the answers to all the questions. And as a result, it made me worried about asking the questions. […] The difference between a junior engineer and a senior engineer is that I know what to search to find the answer to something; I know what to Google or where to look or who to ask, because I’ve already built this weird connection of like ‘I don’t know the answers, but I know how to get the answers’.

RESOURCES

Here are all the most important links/resources mentioned throughout the episode:

Monica’s Website

Monica’s Twitter

Magenta.js

Tenserflow

MidiMe

MIDI City 2000

Yacht’s new album is powered by ML + Artists

 

DON'T FORGET TO SUBSCRIBE!

Help us reach more creatives! Don’t forget to subscribe, review & share – it goes a long way!