On April 19th, 2018 I attended Uber Tech Day 2018. I had such a great time and learned so many wonderful things that I decided to share some of my thoughts and notes from the day.
Why I Went
A good friend of mine sent me an invite to Uber Tech Day 2018 (thanks, Tali!), an event sponsored by LadyEng, their women in engineering group. I decided to go for a number of reasons, some better than others
- It was hosted at a convenient time
- It was free
- It was easy to get to
- It was an opportunity to learn more about a major consumer facing product (as opposed to the industrial products I see in my work at GE)
- It was an opportunity to see how Uber runs events and to see if any of their practices would work for Developer Relations events I am involved in
Who I Heard From and What I Learned from Their Speeches
There were a number of incredible speakers throughout the day. The folks below were not the only speakers I heard, but they were the ones I had the most notes on, and therefore had more lasting takeaways that could be coherently written about than some of the other speakers.
Raquel is a researcher at University of Toronto and is also the head of Uber ATG in Toronto. The goal of her research is to create scalable solutions for autonomous vehicles.
Dr. Urtasun introduced her talk with the two challenges that prevent scaling in the autonomous vehicle space:
- Cost of sensors
Right now, the sensors that are on top of the self-driving cars and trucks you may have seen use LiDAR, which I was familiar with from its use in other scientific discoveries (in particular, a recent discovery in Guatemala.)
What Is Lidar, Why Do Self-Driving Cars Need It, and Can It See Nerf Bullets?
If you've been in Silicon Valley, Pittsburgh, Boston, San Francisco, or any of the other cities where autonomous cars…
LiDAR is super cool, but it is not the cheapest.
2. Cost to map the world
Dr. Urtasun claimed it would cost $2 billion dollars to map the world once — and that is not including the constant updates that would be needed to reflect our ever changing world. I do not know where that figure comes from, but if you think about what is required to map the world and the different possible ways you can do it (drones, satellites, cars with cameras, putting a camera on your cat and letting them run around the neighborhood…) it becomes apparent quickly how complicated and expensive this problem can get.
She continued to explain five areas that her team explores in order to improve the autonomous vehicle space:
- Perception — how do vehicles “see” the world around them? (LiDAR? Cameras? Other types of sensors?) Learning things like what is and is not a car is easy, but figuring out where one car ends and another car begins (differentiation) is a bit harder
- Prediction — what are the objects around an autonomous vehicle (cars, cyclists, pedestrians, animals) going to do given the current context?
- Planning + Control — what should the vehicle do based on the information it has?
- Mapping — gathering information about the roadways to provide info to the vehicles
- Localization — trying to figure out where the vehicle without using a map but instead using variables like the shape of the road, the speed limit, the position of the sun and the time of day, the intersection type
It sounds like Dr. Urtasun and her teams at University of Toronto and at Uber ATG are definitely keeping busy, and I am very curious to see how things evolve as more and more discoveries are made.
Celina Mikolajczak is a super duper battery expert who came to Uber from Tesla in early 2018 to lead battery development for Uber Elevate, Uber’s flying taxi project. The goal currently is to start test flights for it in 2020, and to begin commercial flights in 2023.
Ms. Mikolajczak introduced her talk by talking about expanding urban transport into the third dimension stating that “infrastructure requirements are pretty modest.” This is an especially relevant statement in the Bay Area, where I live, and where expanding BART has been plagued with infrastructure delays. I am a self-described non-early-adopter with some mild to moderate skeptical tendencies, and I also had a wonderful internship years ago in the air traffic control space that left a major impression on me. This means that when Ms. Mikolajczak made this introduction, the next note in my notebook was, “GOOD LUCK.” Though skeptical about the entire concept, I was still eager to hear the rest, as this is a fascinating engineering problem.
These aircraft will need to take off and land vertically, but will travel horizontally. There are some aircraft that exist today that can do this (helicopters included) but none currently on the market meet all of the specifications Uber wants. They will be entirely electric, which is where Ms. Mikolajczak’s battery expertise comes in. These aircraft will only be taking trips in urban areas between “skyports” and these trips will always be less than 60 miles — this is lower than the range of a typical Tesla today. However, the energy costs of the vertical takeoff and landing are rather expensive, and batteries today, while close, are not quite at the capacity needed. The intent is that Uber Elevate aircraft will be charged while they are loading and unloading, as well as when not in use. Right now, the best Tesla battery can be recharged fully in about 45 minutes. The goal for these aircraft is to have those batteries be charged to full capacity in 15 minutes. Today’s battery technology cannot be charged that quickly, but it sounds like it is well on its way.
Ms. Mikolajczak is not tackling this problem alone — she is part of a team of many experts, including folks who have 20+ years of experience at NASA, in the US Military, and in the fields of air traffic control and acoustics. They know what scientific, political, and regulatory challenges stand in their way. They are also partnering with a number of companies with expertise to develop the aircraft itself.
While I am still not an early adopter and am a huge skeptic about most things, but the end of this talk, I definitely transitioned from, “GOOD LUCK,” to “Well, if anyone can do it, these folks can,” and I am eager to see what they achieve.
When Dr. Vivienne Ming first walked onstage, I knew she looked familiar, but I could not figure out why. Finally, it hit me — years ago, I had read an excellent Huffington Post article about Dr. Ming’s unique perspective on being a woman in tech. However, I was sorry to realize that I recalled very little about her actual work, and challenged myself to pay extra attention as a result. It turned out that paying extra attention was no challenge, though, because hearing Dr. Ming speak about her work is captivating. She shared a number of anecdotes, but the most poignant was about the experimenting she has done on her own son.
A few years ago, Dr. Ming’s 6-year-old son was diagnosed with Type 1 diabetes. Since Dr. Ming and her wife are both scientists, they decided to collect as much data as possible from their son. They not only monitored his blood glucose levels, but also monitored every other aspect of his life (what he ate, how he slept, daily activities, emotions) and kept track of it on a Google Doc. When they proudly showed up to their son’s next doctor’s appointment with a stack of papers in hand, they were shocked when the doctor was not excited by the amount of information they had collected. In fact, Dr. Ming says the medical staff was even annoyed at the one inch stack of papers she had brought in, citing it as unnecessary.
Dr. Ming left the appointment feeling, “well, if they don’t want the data, I will just figure this out myself!” And she did. Dr. Ming put a basic smartwatch on her son and continued to monitor all of the data points she had been monitoring previously, and routed all of the data to her private server. She fit this data to a mathematical model that could correctly predict when his blood glucose levels were going to be outside of normal thresholds, and wrote a program to send alerts to a Google Glass that she wore. She later donated this code to Tidepool, which is an open-source not-for-profit company focused on making diabetes data more accessible.
Not only is this research incredible, but I found the reasons behind it to be so simple and so pure. Dr. Ming did not begin this because she woke up one day and decided she wanted to make life easier for folks with diabetes, she began this because she is a mom, and she wanted to help her son. Her passion is inspiring and she is proof that passion, hard work, and resourcefulness can solve problems that make the world a better place!
Dr. Ming has even more research that I won’t discuss here, but if you want to learn more about what she works on, check out Socos Labs, which she (and her wife, Dr. Norma Ming) co-founded to “explore the future of human potential.”
I heard from a number of other speakers from Uber throughout the day, who left me with the following random tidbits from their speeches:
- “If you aren’t planting the seeds for 5–10 years out, then you have no company in 5–10 years” —The Innovators Dilemma
- Context and perspective are crucial in UX. Focus on market fit and also on empathy — you are building for humans, so do your homework to understand them.
- One week of traffic data at Uber is about 1TB. This is about the same amount of data that a single jet engine creates in a cross country flight. Consumer data and industrial data really operate on different scales!
- My algorithms teacher was right — Djikstra’s algorithm is pretty important. Not too much of a surprise given its purpose, but it plays a crucial role in solving the routing problem for Uber.
This day far exceeded my expectations. I came in expecting to hear some interesting technical talks, but did not expect to walk away feeling so inspired that I wanted to write about it and share everything I learned with the world. Thank you, LadyEng and Uber, for putting on such a spectacular event.