Andreas Burg: taming imperfect chips

- EN - FR
’I’ve always loved playing with circuits,’ says Andreas Burg.
’I’ve always loved playing with circuits,’ says Andreas Burg. ©Alain Herzog/ 2020 EPFL

Science as a vocation (1). Andreas Burg is the head of EPFL’s Telecommunication Circuits Laboratory (TCL). He’s a keen proponent of approximate computing, a disciplined approach that prizes "good enough" over perfection.

Ask most scientists what matters to them, and they’ll tell you they strive for perfection, set exacting quality standards and reject rough-and-ready solutions. But Andreas Burg, an associate professor of electrical and electronic engineering, doesn’t fit the mold. For Burg, "good enough" will do. His attitude is perhaps surprising, though not necessarily in a bad way, given that he works on error-prone chips - the most delicate components of modern-day electronic devices. His stance is at odds with conventional thinking in electronics, a discipline normally associated with precision. But over his ten-year career at EPFL, he’s built up solid expertise and become a go-to expert in the fast-growing field of approximate computing.

"Like every scientist, I was taught to assume the worst and to strive for consistent quality," says Burg. "But I chose a different path. For me, things don’t have to be perfect every time. There has to be a trade-off between perfection and other aspects like energy efficiency." Imagine you’re parking your car and you set about making detailed calculations, trying to work out which trajectory will make best use of the available space. It would be a time-consuming process. Now imagine how much time you’d save by just driving into the space without over-thinking it.

The same principle applies in electronics. "Sometimes, good enough is all we need," says Burg. Imagine you’re watching a video on your smartphone but your battery is low. You’d be perfectly happy sacrificing streaming quality if it meant saving energy and preserving your battery life." Embracing imperfection also means accepting minor errors, delays and malfunctions as long as the device does the job it’s designed for.

Makeshift TVs

Burg may accept errors as part of his work, but he doesn’t seem to have made many of them in his career. Growing up in Germany, he was fascinated by electronics from a young age. A confident free-thinker, he moved to Zurich at the tender age of 16. And he’s seized every opportunity that’s come his way ever since. "I’ve always loved playing with circuits," he says. "I remember finding two old TV sets, taking them apart and trying to rebuild them." When the time came to go to college, he couldn’t decide between electrical engineering and physics. So he signed up for both at ETH Zurich. "On my first day, I decided to attend my electrical engineering class," he recalls. "I found it more hands-on than physics, so I stuck with it."

That decision shaped the direction of his career, but it hasn’t always been smooth sailing. "I took six years to finish my thesis," he explains. "Not because I was lazy, but because I was working on so many things at once - communications, cryptography, image and video processing, and other things. Most of that research didn’t make its way into my thesis, but it was fascinating all the same. I knew at that point that I wanted to stay at the university and retain the freedom to explore other aspects of my chosen discipline."

After completing his postdoc, Burg faced another decision: remain in academia or pursue a career in industry. He founded a startup with two colleagues and gave himself two years to make a success of it. "It didn’t work out," he explains. "After two years, we ran out of money. We founded the company to develop WiFi chips. We were pioneers, but we faced stiff competition. We realized that you need more than a chip to build a successful business. It was still a worthwhile experience. I learned a huge amount, and I have no regrets."

Undeterred by the setback, Burg applied for and was awarded a Swiss National Science Foundation (SNSF) professorship in 2009. Two years later, EPFL published a vacancy for a tenure-track professor in telecommunications circuits. "Things were going well at my lab, but I’d been at ETH Zurich for 15 years," he says. "I would’ve been happy to move to the US, but I was lucky enough to land the job at EPFL."

Taking research in radical new directions

Burg is known for his radical thinking, and for having several projects on the go at once. He’s keen to share new ideas and put them to the test in his lab. But he’s equally keen not to waste resources. Ideas with a slim chance of success are quickly abandoned, with only promising concepts getting extra time. He takes a steady, iterative approach to his research, including publishing papers. He developed his main research interest, integrated circuits, during his tenure-track years. But he still retains a passion for other, related subjects. Outside the lab, the concept of approximate computing is gaining ground in industry. "There was a lot of pushback against the concept to begin with," says Burg. "Over time, we’ve learned a lot about industry concerns. Sometimes it’s a case of two steps forward and one step backward, but we’re moving in the right direction."

Burg’s appointment as associate professor has given him greater freedom. "I can be more radical and take more risks in my research," he says. His ambition is to combine communications and artificial intelligence. "Communications is all about equations and mathematical problem-solving," he explains. "Once you introduce AI into the mix, you can start to solve those problems also intuitively."

Burg is also planning to take his research into imperfect chips to the next level. "We’ve developed concepts and predicted how the chips will work," he explains. "The next step is to take these ideas from the drawing board and actually build our own chips. Our initial designs were relatively straightforward. Now we can afford to be more adventurous. EPFL is the perfect place to push the boundaries."

1975 : born in Munich
2000 : completes engineering degree at ETH Zurich
2006 : completes PhD at ETH Zurich
2007 : co-founds ETH Zurich spin-off Celestrius
2009 : obtains SNSF assistant professorship
2011 : appointed tenure-track professor at EPFL
2018 : appointed associate professor at EPFL