Weekly I/O #56
Goodhart's Law, Write More but Less, Annual Report Framework, Claasen's law, Philosophical Jokes
Hi friends,
Greetings from Sunnyvale!
It’s a busy week, but I tried not to skip this entirely for a week. So here's your late weekly dose of I/O. I hope you enjoy it :)
Output
Here's what I've published in the past week.
1. Weekly I/O #55
Last Weekly I/O about Visual before Context, Lossy LLM, Motivation-hygiene Theory, Desirable Difficulty, No More Time.
Input
Here's a list of what I'm exploring and pondering on this week.
1. Goodhart's Law: when a measure becomes a target, it ceases to be a good measure.
Article: GOODHART'S LAW
What might happen when a government offers a bounty on cobra skins because they want to reduce the cobra population? The cobra population increased because people started to breed cobras for their skins and earn money from the government.
When using a metric to reward performance, we provide an incentive to manipulate the metric to gain the reward. Goodhart stated that as: "when a measure becomes a target, it ceases to be a good measure."
The phenomena resulting from this law are ubiquitous because measures of performance (MOPs) are easier to define and qualify than measures of effectiveness (MOEs). However, MOPs are also easier to manipulate, leading to actions that improve performance measurement (cobra skins) but paradoxically reduce the effectiveness (cobra population).
Here's another example. Measuring an engineer's effectiveness is more challenging than measuring their performance based on metrics like the number of lines of code written. If solely the latter metric is used to determine an engineer's promotion, all engineers will strive to produce more lines of code, which could make the quality of code worse and harder to maintain.
2. Write 5x more but write 5x less.
Article: Write 5x more but write 5x less
Mike Crittenden says there are two things he has come to believe about writing:
The average person should write 5x more stuff than they do.
The average written thing should be 5x shorter than it is.
This is such a brilliant way to put it.
3. Farnam Street's Annual Report Framework.
Article: Personal annual report
It might be too late to talk about new year's resolutions, but it's better late than never. I found answering these questions in the list curated by Shane Parrish from Farnam Street helpful because it better shapes the context of any resolution. Here are some I think can be useful for you:
What do I want to spend MORE/LESS time on next year?
Where am I waiting for other people to make the first move?
What can I do in the next week that will make the rest of the year easier?
What can I do this year that will leave me in a better position for next year?
Who do I spend time with that pulls me down/lifts me up?
If there was a film crew following you all day, what am I doing that I'd want the crew to see/avoid them seeing?
If I took over my life from scratch today, what would I immediately stop doing?
4. Claasen's law: Usefulness = log(Technology)
Article: It’s elementary
Theo A. C. M. Claasen, CTO of Philips Semiconductors in 1999, quoted this Logarithmic Law of Usefulness, where usefulness is defined as the perceived value of a new technology compared to its alternatives, measured by its daily-life importance, ease of use, and entertainment value.
Classen argues that a linear improvement in technology's usefulness requires an exponential amount of technological advancements. He gives examples like system performance improves noticeably only if we add memory in increments of 10 or 100 and the bandwidth required for video calls compared with audio calls.
This law is a useful heuristic for evaluating new technology. No matter how magical a new technology is (said self-driving car, AR/VR, LLM), it takes exponential advancement to translate the magic into something useful.
5. “A serious and good philosophical work could be written consisting entirely of jokes." - Ludwig Wittgenstein
Quote
That's it. Thanks for reading. Since I always want to know more about my readers, please let me know which input you find most useful or interesting. You can take 5 seconds and reply to this email with a number!
As always, feel free to send me any interesting ideas you came across recently!
Looking forward to learning from you.
Best,
Cheng-Wei