Nine things I wish I had known the first time I came to NIPS

(This is an approximate transcript of my opening remarks at WiML 2017.)

I’m thrilled to have the honor of kicking off WiML 2017, the 12th Annual Workshop for Women in Machine Learning. I’d like to start by congratulating the organizers who did all the hard work to make this event happen. Let’s give them a round of applause.

These opening remarks often start with a bunch of stats looking at the number of women in different roles within the machine learning community, and examining how these numbers have evolved over the years.

I thought about starting my talk that way, but I couldn’t bring myself to do it.

You have no idea how many times I’ve sat at my computer counting up the number of women on program committees or on boards or giving keynotes at various events. I couldn’t do it again. Are things getting better? Maybe, though — spoiler alert! — if you come to the NIPS opening reception tonight you’ll see a few stats that are as depressing as ever.

Instead I want to focus on the good… on all of you and the fact that you are here with me in this room.

I want to dedicate these opening remarks to the women in this room who are at NIPS for the very first time. And so instead of giving you a bunch of history or statistics, I’m going to use my time to tell you what I wish I had known the first time I came to NIPS, back in 2005.

1. Everyone at NIPS has felt like an impostor at some point. Most people probably still do.

Let’s start with the most important. Machine learning is getting super trendy and there are now all these rock stars and machine learning legends. You, in contrast, just got started and feel like an impostor in the community.

I can tell you this with absolute certainty: Everyone at NIPS has felt like an impostor at some point. Really. Everyone. Even the biggest names. And most people here probably still do, at least some of the time.

Let me tell you a couple thoughts I had while preparing these opening remarks:

Thought #1: Maybe this is a mistake. Nobody is going to want advice from me because I’m not successful enough, or at least not famous enough.

Thought #2: Nobody is going to want advice from me because half my research isn’t even machine learning. Am I even a machine learning person? Why am I here?

Ok, I have been coming to NIPS for 12 years, I’m the tutorials chair this year, I co-founded this very workshop, and I still worry that people will think I don’t belong here. But it’s not just me. We all feel like impostors.

2. It seems like everyone knows each other already. They don’t.

Walking around the conference, it’s easy to get the impression that everyone already knows each other. They don’t.

NIPS is growing super fast and tons of people are new here just like you. Many of them would probably be thrilled if you introduced yourself and said hello. Many of us who have been around for a while would be thrilled to meet you too.

3. You don’t need to go to all the talks. That’s not really the point.

You don’t need to go to all of the talks. In fact, please do not try to go to all of the talks! That’s not the point of a conference.

You’re here for the conversations, for the dinners and the spur-of-the-moment discussions in the hall.

Someone recently told me the “rule of five:” you aren’t going to be able to absorb more than five talks in a day anyway, so don’t force yourself to sit through more.

Frankly, sometimes even five is too many.

4. Don’t just attend the talks you think you should.

Next, you might think that if you’re only going to attend a few talks a day, you should pick the talks most immediately relevant to your own research. But I’d encourage you to use some of your time here exploring one or two topics that are entirely new to you, topics that just sound really exciting. You might find interesting connections to your own work that you never would have expected!

The workshops are particularly good for this since you can fully immerse yourself in something new for a day. Tutorials are great for this too.

5. Asking a question is a good excuse to start a conversation.

Ask speakers questions after their talks. If you’re too shy to do it in front of the whole room, find them in the hallway after. It’s a great excuse to introduce yourself and strike up a conversation.

6. Practicing your research pitch is important, even if it’s painful.

Next up, practice giving your research pitch, even if you don’t enjoy doing it. People you meet are inevitably going to ask what you work on, so why not be prepared?

Keep it high level. Focus on your goals, not the details. Think about why you got excited about this line of work in the first place.

Me? I’m interested in problems that involve interactions between humans and algorithms. My background is in both machine learning theory and algorithmic economics, which deals with algorithm design in contexts in which incentives or strategic behavior come into play. I’ve done a lot of work on prediction markets and crowdsourcing more broadly.

Over the past year I’ve gotten really excited about designing interpretable machine learning techniques, which is super important to think about if we want to get people like doctors, judges, or CEOs trusting machine learning predictions enough to make important decisions. I’ve been collaborating with an interdisciplinary group of colleagues at Microsoft Research to run behavioral experiments aimed to understanding what factors make a model more or less interpretable in different contexts.

I can tell you that I hate doing this. It’s really scary! I have steered so many conversations away from my own research interests over the years because I was afraid of being judged.

Do not be like me! Learn to pitch your research.

7. The most important connections you’ll make are with your peers.

I know everyone wants to meet the rock stars of machine learning, and I’m not discouraging you from trying to do this. But there are a lot of other people at this conference too, and I’d argue that the most important connections to make are with your peers — other grad students or researchers at the same career stage as you. These people are going to be your colleagues for years to come, and some day they are going to be the big names themselves.

As a grad student, I made a handful of amazing friends at NIPS who are now faculty at top universities or researchers at top labs. We still turn to each other for support and career advice all the time. In fact, I’m having dinner with a group of them this week and I’ve been looking forward to it for ages.

I’ll also mention that this very workshop actually started because of friendships made at NIPS…

8. One new friend will often lead to many.

Luckily, one new friend will often lead to many. Cool people often know other cool people. Use this to your advantage. Go along to dinner with new friends that you make and have them introduce you to their network.

You can repay the favor in a couple years when you have your own whole network of friends to introduce.

9. It’s ok to hide and take a break. (Introverts, I’m looking at you.)

Finally, this last one is for all the introverts in the room. You know who you are.

Nobody who knows me will be surprised to hear me say that I am an extreme introvert. I can stand up here and say embarrassing things about myself in front of however many hundreds of people because I’ve trained myself to do it, but there’s a good chance that when this is over I’m going to need to sit by myself and stare at the wall for a while to recover.

And that’s ok!

NIPS is a huge, intense conference, and it’s ok to hide and take a break when you need to. Read a book, go for a walk along the beach, get some exercise, whatever you need to do. Be aware of your own needs and don’t feel guilty about them.

With that, I just want to take one more chance to say welcome to Women in Machine Learning! Make the most of your time here. Talk to each other, ask lots of questions, make new friends. Practice your research pitch on other people in this room today, and enjoy the week at NIPS!

Senior Principal Researcher at Microsoft Research, New York City