top of page

Interview: Dan Tepfer explains how he jams with a computer on stage

The Brooklyn-based jazz musician brings his "Natural Machines" show to the NYC stage this week. Click here to read full article on amNewYork website.

Live jazz lives for improvisation. Watching a seasoned trio of players go off on their own flights of fancy while staying together in the same song can be, in an era of buttoned-down and over-choreographed concerts, one of the most interesting experiences for music lovers. But what happens when one of the players is a set of algorithms?

When jazz pianist Dan Tepfer realized that the Disklavier — a self-playing piano often seen in hotel lounges or airports — could not only play, but also take in real-time input, he had an idea: what if he could set up the Disklavier to respond to his playing, while he was doing the same thing? It’s produced an album, “Natural Machines,” and a touring show that features Tepfer improvising off a Disklavier set up to not just play alongside him, but also “respond” to his playing via an algorithm (Tepfer wrote all of the programming himself). The result is a computer-human hybrid that takes on the character of the latter much more than the former.

Brooklyn-based Tepfer, 37, spoke with us in advance of his show at National Sawdust about improvisation, programming and virtual reality.

How different is it to improvise off a computer rather than a human player?

In many ways, it’s less different than you might think. It’s a little bit like the concept in improv comedy of a “yes, and.” In both cases you have to accept what your improvisation partner is doing and run with it. My biggest focus when I’m performing the “Natural Machines” program is to just really be listening as carefully as I can to what the computer’s doing, and to be engaging with it in the moment, which is exactly what I try to do when I’m improvising with a person. The big difference is that the computer is responding to rules, and a person responds to something a little more mysterious.

Were there rules for writing the algorithms? Was there some programming that just didn’t work?

The genesis of this project is me writing an algorithm, thinking “this could be fun,” and then just playing with it and seeing where it takes me. Some of them work and some of them don’t work as well. But I might find a way to make the ones that don’t work as well work by playing a different way. That’s where this gets interesting to me: You set up some rules, because that’s what an algorithm is, and then how do you work with those rules. So, in some ways, the question isn’t “do I need to write the algorithms a certain way so that they sound good?” and more “do I need to play a certain way with an algorithm in order to make it sound good?”

So, it’s more in the adaptation of what you would do as a player than what you would do as a programmer.

That’s where it gets interesting to me — “Let's come up with some wild rules, and then let’s see what happens. Let’s see if I can make it work.” Because that’s what we do as human beings. We’re adaptable. But we have to set up situations for ourselves in which we need to adapt, otherwise we end up doing the same thing all of the time.

The visuals for the show, including virtual reality elements, are triggered by your playing as well — so everything is essentially being improvised?

One of the essential aspects of this project is that everything is real time. Everything you see in the graphics, that way the computer is responding on the piano — nothing is preplanned. Only the rules are. That’s also true of VR, so you can imagine there are some pretty serious challenges there — you’ve got to get the data not only from what I’m playing but also from what the piano’s playing, to all these phones in real time with a minimal latency. But I figured out a way to do it.

bottom of page