2020: (half) year in review - Democratizing Automation
How I ended up with 26 posts in my first half year of weekly writing. What we learned and where we are going.
This project had its first post on June 12th, and in total has seen 26 posts (I didn’t plan that for perfectly being a half year), but it worked out!
I am sending out my predictions for 2021 in automation and AI soon, so this will kickstart the next phase of our journey. This post is a little review of what we have learned about this year, and can serve as a starting point for new or recent joiners!
Most Viewed Articles
Predicting view numbers has been a total wash with a small readership. I say for every two projects I am very excited about, one of them gets the traction I expected. Maybe my expectations got pulled too high after my first article was much much more viewed.
10 years of automation in 1 year: my first article on this site has gotten 4x the views of any other. It was the start of my journey into figuring out how COVID19 is changing the adoption of automation in different sectors of the economy. This will be revisited shortly in 2021, so keep an eye out for it.
Recommendations are a game - a dangerous game (for us) (author fav.): this was my first blog project — try to dissect why automated systems that act in a broad way on our society are difficult to model (and connected to a recent line of work on reinforcement learning policy).
The uncanny world of robots at home (author fav.): I did a little digging into the uncanny valley (as part of a series on re-thinking robots that is on pause) and found a ton of fun facts in the history of the uncanny, or should I say bukimi no tani genshō.
Online courses, automating education, and digitalizing degrees: a tour of how the pandemic is changing (and ruining) plenty of our educational norms.
Models, systems, code; and robots (author fav.): for a long time I have tried to work out the different ways mechanical engineers, computer scientists, and electrical engineers view problems — this was my hypothesis in writing.
A lot of the trends I hit on this year came up three or four times. It’s a lot easier to see where these are going when presented in the order they appeared.
Education and Digitalization
The pandemic has drawn on longer than expected, giving us a little more time to see the changes happen before our eyes (as opposed to at a slower rate in the coming years, started by COVID). The pandemic is keeping the pedal on the floor for changes to education tools. I suspect next year we see a university purely digitize its degree for a lower price-point to tap into an unaddressed market: name recognition without the experience.
Online courses, automating education, and digitalizing degrees: the first reflection on what the pandemic is doing to courses at UC Berkeley.
Why Ph.D. students are lonelier than retirees: a reprint of the problems COVID is exasperating with respect to the mental health of graduate students.
Some thoughts on teaching & learning: a second reflection on how students make things more difficult for themselves — in terms of learning the material — when courses are online and asynchronous.
Algorithms and Everyday Life
When I started this blog I stated a long-term goal of “figuring out how automation will disproportionally affect lower-income people who cannot afford human treatment nor are educated on the risks of their digital lives.” This reflection started with myself, so I studied how the different algorithms affected me. Eventually, I started getting into bigger trends like what defining truth would look like and how Facebook uses AI in its product safety. This is likely the central topic of my blog and will definitely be revisited.
Recommendations are a game - a dangerous game (for us): read this if you want to be scared of the downstream effects of RL at scale.
Automated: how algorithms shape a day in the life and our future: why I am criminally addicted to Twitter and not some other websites.
Facebook case study 🌏🔬: using AI to regulate the Digital ecosystem: Facebook’s lukewarm handling of hate-speech and misinformation.
AI & arbitration of truth: how Trump’s craziness is changing what our social networks publish.
Robotics and Automation Policy
With me joining an AI Ethics group at UC Berkeley (plug for GEESE), I started formulating more thoughts on what some of these parallels mean for roboticists. In short, it is more difficult because robotics is broadly any automation that interacts with the physical world. This is a big work in progress, but something that could grow into a really useful portfolio of knowledge.
Towards an ethics for roboticists: why creating an ethics guidebook is so hard.
Constructing axes for reinforcement learning policy (author fav.): addressing the slipperiness of RL and why creating policy on it will be so hard.
The Collingridge Dilemma and current policy on robots (author fav.): a deep dive into what legal precedent there actually is regarding robotics.
What is a Robot?
Robotics is really determined by science-fiction, so I enjoyed spending some time figuring out what a robot should mean in the future. We need a broader definition, else our robots will not be accomplishing broad tasks.
The uncanny world of robots at home (author fav.): why consumer robots are going to cause more harm than good in the short term.
Does autonomy = intelligence?: what does it mean to create an intelligent agent in the home?
Digital companies and the goal of at-home embodied AI (author fav.): explaining why Facebook, and hopefully other tech companies, have employed me.
Autonomy startups are such a mess: explaining why I am reluctant to join most robotics startups.
These have been archived to have a cleaner feed for new visitors.
My sub-newsletter, Tangentially Related is where I put all of my thoughts on current events and other interests (longevity, maybe sports may creep in, my research, not politics, etc.). The occasional recap on robots and automation ended up being posted about once a month, see August, September, October, and November.
The common themes for these this year were consumer robots, mental health & longevity, and tech company snippets. I am happy with these posts, but let me know if there is anything you would want to see added or removed from them. Personally, the best part of writing these is the “Snacks!” section at the end.
What is next?
This newsletter now has over 160 subscribers and is growing (slowly) consistently via random internet traffic. I want to try and turn this into more of a community where people can engage, learn from each other, and create more value. If you want to be a part of this for next year, send this to someone and tell them to subscribe.
I appreciate all of the very direct positive and constructive feedback I have got. For more of the new followers who I do not know, my DMs on Twitter are always open (or you can reply to this email or comment). Also, you can follow this blog directly on Twitter instead of email. Forwarded this? Subscribe below.