What is the essence of Paul McCartney?
And if you can work that out, could you distil and then sell it?
It’s a crazy idea, but François Pachet thinks it’s possible.
He’s the director of Sony’s Computer Science Laboratory Paris, where he leads the music research team. And he believes that the marketing of musical ’style’ – as opposed to recordings – could herald the next commercial opportunity for the music industry and musicians.
Pachet is uniquely qualified to pursue such an eccentric mission. He is an accomplished jazz musician who regularly performs and has even recorded an album. Yet he is also a computer scientist.
Indeed, he has a PhD in artificial intelligence.
At Sony, Pachet is developing authoring tools to ‘abstract’ style and ‘turn it into a ‘malleable substance that acts as a texture’.
If it sounds bizarre, that’s because it is. But Pachet is a tech industry moonshot thinker, an employee given free rein to ponder the future and come up with wild ideas.
In recent weeks, Hot Topics spoke to Intel’s moonshot thinker Steve Brown as part of an occasional series.
Here’s what happened when we sat down with François Pachet…
What led to your recruitment by Sony Labs?
I was a musician and a computer scientist. I’d been to Berkeley and I’d done my PhD in Paris, and then became assistant professor. But 17 years ago I met the people at Sony. They had created a Computer Science Laboratory in Tokyo and they wanted to do the same in Paris.
I was the right guy in the right place at the right time. It was great. I could do the research exactly as I wanted to do at Sony. I had more freedom and more resources than I had at the university.
How do you summarise this research?
Our end goal is to create tools that will help people be more creative. The idea comes from looking at great creators like Picasso.
They were very good at reproducing the style of their predecessors, and would then take a style they knew well and bring new ideas to it.
If you look at this mathematically, it’s a very difficult combernatorial problem. And that’s what we’re trying to solve with the Flow Machines Project using something called a Markov model.
The idea is to have a machine do the computation so that artists can navigate around these challenges without having to personally solve them.
How might that work in practice?
Let’s say you want to compose a melody in style of Paul McCartney and you want to start with these notes and that chord. The system will understand your constraints and create a melody for you.
Afterwards, you can review it and change things, and have a dialogue with the machine. The important thing is to give musicians tools to try things out.
I think people would pay for musical style if you could distil it.
But isn’t this kind of experimentation what good musicians do anyway? Why do they need machine assistance?
A lot of musicians are frustrated all the time. Great artists like John Coltrane and Charlie Parker had stuff in their heads that they couldn’t reproduce. This can be the engine of creativity.
Very often you’re trying to find a solution to a specific problem and the computer can help you find it.
Do you think being a musician yourself – especially a jazz musician – makes you better able to be a moonshot thinker, and to be creative in other areas?
Yes. I’m convinced that being able to play jazz has an impact on how you think and find solutions. It’s because of the decision-making and improvisation involved.
You have to deal with constraints in terms of harmony and rhythm and arrangement yet find something interesting to say.
How easy is it to be a moonshot thinker while also being the manager of a department?
The challenge of running the lab is more about finding the right people to work with. We’re looking for people with obsessions and who are not afraid to say what they think. They need to be crazy, but not too much.
They also have to be selfish in the sense that they are consumed by their work, yet able to work with other people.
Perhaps the hardest thing is to identify people who can resist the distraction of email and social media and YouTube cats.
How does Sony benefit from your work?
There’s no direct commercial benefit. And much of the funding comes from the European Research Council.
So how are you evaluated?
I don’t have a formal appraisal. This comes back to the original founder of the lab, Mario Tokoro. He’s not interested in traditional methods of evaluation, so instead I do interviews with artists or give talks and write papers.
We don’t have meetings and committees. Mario hates all that.
It’s such an unusual job. What is your guiding principle for how you approach it?
I’ll reference something Mario said: if you have an idea, ask yourself what happens if you don’t do it.
If you think someone else will do it anyway, then forget about it. You should do only things no one else will do.
That’s so different from the norm…
Yeah, the western way is to do it faster than the other guy. Our philosophy is to look for things that make you different. The fact that it’s singular is more important even than whether it’s good or bad.