Admiration, running, and AI
Which living person do you most admire?
Andrew Herbert. Andrew was the managing director of Microsoft Research Cambridge from 2003 – 2011, and he was the person that taught me everything about applied science. Andrew is one of the few people who combine scientific depth, entrepreneurial spirit (he co-founded his own company 20 years ago) and solid engineering skills. He is one of the most approachable and humble people I know.
What is your greatest fear?
Heights – I can hardly stand three steps on a ladder without feeling very dizzy. And it gets worse with every additional step. I have been known to completely freeze when leaving the elevator of the Eiffel tower – rather than walk down, I have had to take the elevator down again!
In your mind, what is the single greatest innovation in history?
The steam engine – it was the first tool that automated human labor and has changed every industry and everyone’s life. Since the invention of the steam engine, people’s life expectancy has increased, we can travel longer distances, we live longer, and we can use our intellect rather than rely just on our physical strength. The world has become a more prosperous and connected place thanks to it.
What is your current state of mind?
Calm and determined.
What is one thing you do every week you do to stay focused on customer needs?
Every day I get a subset of question-mark emails forwarded to me, and I read every one of them in detail. We (my team) also have collaboration stories in the first minute of all our meetings. These concrete stories of collaboration successes and misses with our internal business group customers help me learn a lot on what is or is not working.
What do you consider the most overrated virtue?
Patience. Being impatient can be a bad thing in scientific experiments. However, in general, being impatient is what moves the whole team forward.
What publications do you read to stay current with external trends?
On the scientific side, I read every single submission authored by Amazon scientists. That’s over 200 per year so this is nearly one paper per day. On the technical side, I read TechCrunch and Wired but rely on friends or news services to push articles to me.
Which words or phrases do you most overuse?
“Mechanism.” I am a big believer in organizational mechanisms to allow leaders to scale, and I find myself too often asking, “Do you have a mechanism for that?” or “What mechanisms are you using in your team?”
What or who is the greatest love of your life?
My wife, Jeannette Herbrich. I met her when I was 19, and we have shared so many things together over the past 25 years. We have lived in three different countries, and I cannot wait to see what will happen in the next 25 years.
When and where were you happiest?
I love every second of what I am doing at work and often need to restrain myself not to work all the time. Outside of work, the happiest moment was when I finished the (authentic) marathon in Athens last year – the emotions that one feels after four and a half hours of running through this scenic route with lots of history are hard to describe in words. You experience a heady mix of accomplishments, euphoria, and exhaustion.
What was the last time you were wrong about something?
Just a couple of weeks ago – we were discussing what operation in an ML hardware system consumes a lot of energy. I had assumed that data movement and holding state in memory would be expensive operation. However, for deep learning systems, it turns out that the energy consumption is no more than 20%. I think it’s totally fine to be wrong about something as long as one makes sure to not repeat the same mistake again. I learn the most from my mistakes so I will never forget the above fact now.
Which talent would you most like to have?
I wish I could draw better. I am a very visual person and I use sketching a lot at work to explain scientific concepts. But I cannot draw scenes with humans or animals in any way that it is recognizable.
Tell us about the last time you disagreed and committed to something.
I do this a lot – my team members are the best scientists in their field, and scientists disagree all the time! One time I remember is a debate we had about what constitutes a high-impact application of energy-efficient machine learning: I believed it was the ability to play a board game (e.g. Chess or Go) better than humans with as little energy as a human player. But the team said that this is not only harder, but also not derived from actual energy-efficiency problems that we have in the machine translation space. I committed and the team was right.
If you could change one thing about yourself, what would it be?
To take more time for reading non-technical books. I consume tons of technical literature but the number of fiction and non-fiction stories I have read are few and far between. I enjoy reading biographies and books about running and food. However, I hardly prioritize it high enough to “just read” for pure enjoyment. That only happens when I am on a long vacation.
What do you consider your greatest achievement?
My two children – they are 13 and 16 and I am very proud how independent and happy they are in life and how many friendships they have made.
If you were to die and come back as a person or a thing, what would it be?
I would happily come back and live my life again – there has not been a single dull moment and I have no regrets that I missed out on anything.
Where would you most like to live?
I lived in Cambridge (UK), Mountain View (US) and Berlin (Germany). To be honest, I am extremely happy where I am living right now. However, the one country on the wish-list of both my wife and myself is New Zealand. Our plan is to spend one year living there when our daughter turns 18 and leaves the house.
What is your most treasured possession?
I don’t really have a most treasured possession. When I was 15 years old, Germany re-united, and all of a sudden, I was able to travel the world and study abroad (I grew up in East Germany). So much changed overnight, and I have since never grown attached to any artifact. If there is one possession I value highly and have attachment to, it’s my wedding ring; I have worn it every day for the past 18 years and it has lived many of my memories with me.
What are some innovations from your organization in 2017 that inspire the most pride within you?
I am proud of every innovation that the Core ML teams have developed – both the ones that were successful, and the ones that failed and we learned a lot from. But there is one innovation that is highly visible and has been many years in the making: our automated freshness prediction conveyance system (deployed in LAX6). It’s taken 3.5 years from a mere idea to a fully functioning system. It’s really amazing to see this in action now, and marvel at a world where less fresh food goes to waste because of this innovation.
What was your first day like at Amazon?
Chaotic and a culture shock. First, no one had given me the address of the New Hire Orientation event, so I was wandering around Van Vorst (a plaza on the Amazon campus in Seattle). But I found the class just in time. After that, I attended three doc reviews where everyone was reading for half the meeting. I was wondering what is going on here but went along with it. I called my wife that evening and said "This is amazing. They are actually taking time to read papers here at Amazon!"
What is an example of a piece of advice you have received and what is an example of a piece of advice you have delivered to others?
It’s the same advice that I received which is the advice I am giving to everyone at Amazon: You have to learn to say "No" and focus. You have to learn to own your time and be frugal with it because no-one else will do it for you. It took me six months to feel confident enough to say "No, I cannot own this."
What is an example of how machine learning is helping Amazon meet customer needs?
There are literally hundreds so it’s hard to pick one specific one. To give an example, we automatically detect and prevent copyright-infringing content made available on Amazon Video.
What is your favorite sci-fi book of all time?
As I said earlier, I don't read enough non-technical books to have a favorite sci-fi book of all time. I am a movie geek, and my favorite sci-fi movie of all times is “The Matrix” – it’s an intriguing story how mind, religion, and the physical world can be reconciled.
What was the last thing you purchased on Amazon?
Powerbar Gel Original (24 x 41g pack). This is for my weekly half-marathon runs.