Bootstrapping like it's 1908 and the Apollo Guidance Computer
How new techniques are re-hashes of old, about this newsletter, and some fun facts about the Apollo guidance computer.
Hello!
Gratitude for giving me the trust to invade your inbox. Let me know if something is amiss.
New articles
Entropic Thoughts: New Name for Two-Wrongs
I have long wanted to rename Two Wrongs because although the name started out aptly, it has not fit the site for many years. I share a few more details about canonical URLs.
Full article (1–2 minute read): Entropic Thoughts: New Name for Two-Wrongs
Launching a Much-Requested Newsletter
This is it! You have found it. The generosity of Justin Duke at Buttondown made it possible.
Full article (1–2 minute read): Launching a Much-Requested Newsletter
Student Resampled His t Distribution
Back in 1908, Student was an industrial statistician working at Guinness, trying to figure out how to get useful science out of small samples. He eventually derived a theoretical distribution for it, which we now know as Student's t. It is commonly encountered in the small-sample hypothesis test known as the t test.
But! Before he managed to figure it out theoretically, he actually discovered its shape by experimentation. He wrote out 3000 measurements on as many pieces of paper and shuffled them in a hat and then drew them one at a time and wrote them back into a table – just to randomise the order of the table he got them from!
That's inspirational to me. If Student didn't figure it out theoretically at first, then maybe also we don't have to.
Also some observations on computational power.
Full article (2–3 minute read): Student Resampled His t Distribution
Flashcard of the week
One of my flashcards asks
What does AGC stand for?
This one is easy for me: Apollo Guidance Computer. It was the first digital computer used in an important project. Here are three more fun facts about this computer:
The AGC was so central to the mission that they wanted a backup copy of each of its components going up into space. They realised that they could improve reliability by just having a fully assembled backup computer instead: this let them not worry about connections for swapping components.
Before they decided on carrying backups of all components, there were some discussions on whether the astronauts should be responsible for repairing some components during a mission.
When Shepherd got the instructions for one of those repair jobs, he asked something along the lines of "Why don't we also learn a little brain surgery in case one of us needs to be operated on?" Apparently fixing computers is at the level of brain surgery even for astronauts!
When MIT asked for reliability specifications of the AGC, nobody really knew what to answer. The mean time between failure of existing digital computers (serving auxiliary functions in fighter jets) was about 14 hours, which was well within margins for a fighter jet but clearly insufficient for a lunar mission.
One suggestion came down as "the AGC shall not put astronauts in lethal risk exceeding 0.1 %". Someone at MIT dug out actuarial tables for three forty-year olds and it turns out that during two weeks in space, that was about the rate of fatalities that were expected of common causes alone!
The fact that the AGC was the first important digital computer had lots of interesting consequences for the development of the Apollo missions, as well as leading to some of the earliest observations on software development. I have written about this before, and I intend to do it at least once more.