2019: A Personal Retrospective
It was a year of deep study and contemplation. 2020 will take things further.
Human Factors, System Safety, and Resilience
I opened the year by attending a workshop at Lund University entitled Critical Thinking in Safety. My objective was to more deeply understand how failures and outages come about in software systems and what to do about them. At work (Microsoft) this topic was one of my main concerns, as I was involved in shaping the policies and methods teams were expected to use to improve the reliability of their services. I had done a study on outage pattern analysis the year before, without satisfying results. Our learnings from the study were shared at SREcon18 Americas. I knew I was missing something, if not several things, and set out to right my understanding of human activity in complex, uncertain domains.
At the workshop at Lund my eyes were opened to a large corpus of existing research about similar questions. Though the research was mostly not conducted on software systems but rather other high-risk, complex endeavors such as firefighting, aviation, and medicine, I found many of the principles applicable to my own daily work. I spent an intensive week with interesting people debating and discussing research during lab hours each day and late into the night afterwards. I summarized my notes and initial impressions a this series of articles. I then went to work implementing and sharing what I’d learned among my peers at Microsoft as we sought to strengthen Azure.
Around this time I also joined the “Learning From Incidents” slack group and became better connected with others in software asking similar questions. This group has subsequently launched a website: https://www.learningfromincidents.io/, to which I might get around to contributing at some point.
Real World Risk
The next deep study I made was about real-world risk and decision making at the Real World Risk Institute (RWRI) in NYC. This is an intensive workshop lead by Nassim Taleb (yes, that Nassim), Robert Frey, and Raphael Douady with several recurring and guest lecturers. The focus was on how to understand and take risks in the real world.
The curriculum was of a skeptical empiricist tradition, valuing heuristics that have stood the test of time while questioning nice-sounding theories that haven’t been proven in reality. One major principle that stood out to me was “iatrogenics”, or harm caused by the healer. In complex situations, undesirable side-effects of otherwise nice-sounding theories can do more harm than good. I won’t re-hash things here, but rather point you to my series of articles about the week.
Both Lund University and RWRI helped me to better understand the complexities and non-linear interactions between components in life and in systems (despite the disagreements between these two schools of thought). In my work, it helped me identify policies/rules/metrics that would do more harm than good as well as how to find and reinforce the things already keeping Azure afloat (the human practitioners, usually) while addressing systemic weaknesses.
In my personal life, I was able to think more holistically and realistically about risk. My B.S. detector seems to be stronger too. I see the world quite differently now.
Health & Fitness
My health had deteriorated over the past two years thanks to bad habits, my desk job, and the extra work hours I put in every evening pursuing a graduate degree. After my wife and I welcomed another child into our home, it was time to make things right.
There is a lot of snake oil in the worlds of health and nutrition so I decided to start simple. I got a gym membership and a set of six 1-on-1 training sessions to learn “proper” form and address initial weakness/imbalance. Afterwards, I moved to a simple circuit of heavy weights.
I subsequently found an empirically tested resource (Body By Science) and shifted to a 20 minute, once-per-week, super slow, high intensity training regimen. This has been more than enough to make significant progress in strength, body composition, and energy levels. Non-linear.
SREcon19 APAC
In the summer I traveled to Singapore to present at the opening and closing plenaries for SREcon19 APAC. It was a fun and enriching experience, and I took full advantage of the opportunity to explore a close friend’s ancestral homeland while also discussing complexity, iatrogenics, causality, and risk with others in the software industry.
Frankly, I think the presentations themselves are poorly delivered, but the content is valuable:
Relocation
For various personal, familial, and professional reasons, I ended my time at Microsoft in August and began a new chapter at Adobe.
I’ve been applying my experience with distributed systems, incident management, complexity, and risk to their cloud storage platform. I’m based in the Lehi, UT office.
Reinforcement Learning & Decision Making
This fall, I wrapped up my graduate studies at Georgia Tech by taking two classes: AI for Robotics (taught by Sebastian Thrun) and Reinforcement Learning (taught by Charles Isbell and Michael Littman).
Perhaps unsurprisingly, many concepts overlapped with other things I studied this year. I’ve caught on to something here. In December I completed my studies and graduated with an MS in CS.
Other Impactful Resources (I’ll spare you my fiction reading list)
- Reinforcement Learning — David Silver
- Statistical Rethinking — Richard McElreath
- Behind Human Error — David Woods et al.
- The Tyranny of Experts — William Easterly
- Birth of a Theorem — Cédric Villani
- The Second Book of Nephi — Nephi Son of Lehi
Reflections
I feel that I am starting to tug at threads that tie many different disciplines together. This merits deeper study, and I intend to research decision making, safety, and risk taking further in 2020. There’s something here that I can’t quite articulate yet; a question that remains unanswered and perhaps unspoken. To start, this will likely involve:
- Statistics
- Reinforcement Learning
- Optimization
- Option Theory
- Game Theory
- Ergodicity Economics
- Resilience Engineering
- System Safety
- Embodied Cognitive Science
- Operations Research
- Robotics
- Ancient Wisdom
There will be a mix of study and application (as all learning should have) and I will be learning in the open. Artifacts will be created as appropriate. I’ve no funding for this research at present, so it will fit in between my more important family responsibilities and my full-time employment.
Please contact me on twitter or here if you would like to discuss these topics. Friends, collaborators, and community can have an outsized impact and I welcome unexpected opportunities.
