Thoughts on the essay "Learning to Work"

I and many of my friends have at times taken up meditation practices, or yoga practices, or other daily habits. These usually have a similar structure, where every day, for a specified amount of time, we task ourselves with honing a specific technique, gently building the muscles of the practice so we gradually but surely expand our abilities. In the 1976 essay “Learning to Work” Virginia Valian presents us with a similar structure. Instead of for a hobby though, she applies it to intellectual labor. While for many of us in knowledge professions, intellectual work is our most treasured ability to hone, most people do not set aside intentional time every day to hone their capabilities. I know I haven’t. While I have dedicated myself to dancing, or working out at times before, I’ve never developed a loving, gentle daily practice of increasing my abilities. Instead, I find myself throwing myself at tasks haphazardly, letting my whims take me where they will go.

Read More


Hopefully many of you are out there scheming on how to leverage the most recent advances in AI to create value for the world. Likely you plan on capturing some of that value. Generative models are amazing for supplementing creativity, but outside of generating text, images, and videos, one might be struck with the question about how else they could apply the amazing capabilities of these language models.

Read More

Monitoring for Deceptive AI

In the court of the Evil AI resistance team this week, Evan Hubinger, who is fully employed on the AI resistance team at an institute called MIRI, has come up with a plan he suggests for Big AI to follow, such as Google and OpenAI. He thinks that while these companies try to train stronger and stronger AIs, an easy thing they could do to help prevent the AIs from becoming evil is to watch for early signs of deception coming from the models. One of the ways the resistance is worried that powerful AI might exhibit evilness and badness is to make everything look hunky-dory to the people training it to not be evil, but secretly it’s actually very much being evil. While the AI isn’t so very good just yet at deceiving its trainers, Evan says it should be pretty easy to catch it in the act.

Read More

My experience at a Silent Meditation Retreat

Hello, and thank you for checking out my post. I recently attended a silent meditation retreat that lasted 10 days at a center near Lyon, France. During the time I was there, I finally found a taste of some windows one can look at the world through which I knew must exist, but that I had not gazed out through.

Read More

Artificial Photosynthesis

Over the past two years of reading about the field of artificial photosynthesis, I have become convinced that the field is now at the point where an influx of funding could give rise to a fast tech transfer.

Read More

Building Software Libraries

  • No service owner understands every package in their software.
  • If there is a bug in your library, noticing the regression will be delayed after the release, as you don’t notice until somebody uses that release.
  • Keep fewer codepaths (cyclomatic complexity)
  • think about what other software might be running on service.
  • You can explicitly version within package names to prevent against breaking changes, if you want to be super careful.
  • Clients shouldn’t need to write boilerplate to use your library
  • Libraries should be self-documenting - especially by throwing checked exceptions on edge cases.
  • As a library owner, you have to be aware of any assumptions you’re making about the outside world.
  • Once you release a library, you really don’t want to need to touch it, because you may break people.
  • This means that you have to do a really good job at requirements gathering, and then don’t touch code unless it NEEDS to be touched.
  • Be paranoid about changes to your package.
  • Test your code with actual customers before releasing it publicly.
  • You want to depend on as few packages as possible as a library, this makes it way easier to consume.
  • Be cognizant about the logs and exceptions you spew, definitely log and except, but not too much.
  • Link to wikis inside of exception messages. However, wikis can get out of date.
  • Libraries should be: (Reliable, Intuitive, Secure, Efficient, Maintainable, Universal, Debuggable)
  • It is possible to create a service that is interacted with as a library, which is beneficial in some situations
Read More

What Exactly *Is* a Difficult Decision?

After work today I had on my to-do list to complete a set of short-answer questions. One of the prompts was - what is a “difficult decision” I had to make at my job? I had no idea how to answer this, because I didn’t really know what qualified as a difficult decision.

Read More

Scientific Framing

On this beautiful Sunday morning, I’ve been reading a book from 1976, called Computer Power and Human Reason. The thesis of the book is that computers are a powerful frame intoxicant. He introduces his position with a parable.

Read More


When you think of your vote in the great 2020 election, ask yourself why you voted for who you did. After four years of being barraged by the media, you probably reason you voted for Trump which I know all of you did because of one big thing, such as “the economy” and then maybe if you spent long enough thinking about it, a few other things, such as how much you hate “Black Lives Matter” or how not-with-it Biden seems. Whatever your reason for voting for Trump, I put forward that if you were left to research cold hard facts from unbiased sources, you would find quite a different set of issues as the most prominent, and probably happen to ignore at least one minor reason that contributed to your vote.

Read More

The Solvable Problem and the Real Problem

This election cycle, I tried to get a job at one of the companies that receive the donations you give to your preferred candidate, that try to convince those Other People to vote like you do. I failed to secure one of these jobs. I made it decently far in a few interview loops but ended up with bupkis. For this essay, I am concentrating on the interview loop that was the shortest; you could also refer to it as a “single interview”.

Read More

Voting Trump out of Love

I canvassed in Maine today, two days before the election. Global Warming meant that the trees still carried their fall radiance. I drove up on my own, because I couldn’t manage to convince any of my Boston friends to join me. After I picked up my terf and parked my car by an abandoned physical therapy practice, I was so anxious to start canvassing I had to take a hit from my dugout to get out of the car.

Read More

Autoimmune Diseases

Autoimmune diseases are a category of illness, involving many different mechanisms and causes. The medicines we use to treat them work by interacting with your immune system in a few ways

Read More

Super-Emitters in California (Landfills)

Forward landfill This Saturday I spent some time looking through the Methane Source Finder put out by Riley Duren of JPL. The paper can be found paywalled here. The authors allow that landfills are by far the largest contributor to the issue, but interestingly don’t “name and shame” the locations and operators, which makes sense because they are working with the operators to clear up the issues which is great. However, I of course will name and shame them. Here are the top 5.

Read More

Patellofemoral Pain and Me

What is Patellofemoral Pain Syndrome?

The patellofemoral joint is made up of the fossa where the patella slides between the two femoral condyles. The word patella means a small dish in greek. The word condyle means “knuckle”, and is one of the pair of rounded thing on the end of bones. PFP is also known as runners’ knee or Chondromalacia patellae. The word chondromalacia is derived from the Greek chrondros, meaning cartilage and malakia, meaning softening.

Read More

Making healthcare in ML fair, and measures of how to do it

So this article has been making the rounds recently . The idea is that there are algorithms used to recommend patients for more intensive care which predict using the “label” of the future cost of the patient. While using a label related to cost is easy because the data can be collected simply, and it is a simple numeric value, it embeds the history of black ppl getting less healthcare done, and the paper claims that this means that blacks have significantly higher amount of sickness conditioned on this risk score than white ppl. The question of whether “amount of sickness” is tied to their underlying metric of number of chronic conditions seems open to me, but opening up the conversation of the use of money as a proxy for other labels seems like a useful addition to the public dialogue.

Read More

Things I learned from my Internship

About Myself

  • I work much harder when I feel like there’s work I should be doing, that I know how to start on, and that improves something I care about
  • Starting things is hard
  • I can work all day effectively, if the work isn’t super brain-intensive.
  • I like making cleanliness and ease of understanding a perogative
  • I’m good but fairly slow at programming.
  • I’m slow when I don’t stay cognizant on what I’m trying to accomplish, and the steps I’m taking to get there.
  • I’m slow to make up my mind
  • I need to work on creating a decisive, cleaned-up front to approach business with
  • My memory is pretty crap, I think I remember most things for like 2 weeks. so I should write things down when I learn them.
Read More