air purifiers, democracy by mistake & the social science paradox
CC#76 - Things LLMs Are Not So Good At, Headless Chicken & Egg Smell
Hey there and welcome to ✨ CuratedCuriosity - a bi-weekly newsletter delivering inspiration from all over the internet to the notoriously curious.
Things I Enjoyed Reading.
❌ Democracy by Mistake
How does democracy “happen”? There a few competing theories from political science about what needs to happen in order to make a country become democratic. This article adopts a bit of a different point of view - the claim here is that democracy is ‘caused’ by irrational behaviour i.e. mistakes by the rulers in place.
“I compiled a synopsis of each episode. I looked for observable implications of the various deliberate-choice theories of democratization. Did the political transition follow protests, strikes, or other mass actions? Over what issues? Did democratization occur around a time of war, civil war, or significant threat of these? Did the reformers think newly enfranchised citizens would vote for them?
The more I read, the stronger my hunch grew. There were cases that plausibly fit each of the rationales I had distilled from the explanatory literature. Just not that many. (…)
Myopia and miscalculation are far more obvious in contemporaneous accounts than careful planning. Even when not swept from power unexpectedly, leaders often drifted into reform because they found themselves cornered by their own previous moves. Democracy emerged as a result of their errors. Indeed, in the vast majority of historical cases, the evidence fits a scenario I call “democracy by mistake.”
💨 Better air quality is the easiest way not to die
I have been cycling around Cambridge quite a lot and while there are many bikelanes, the majority of them are just on the side of big streets (often with two lanes). This got me thinking about air pollution, which in turn made me stumble upon this article. While its a bit hard to figure out whether cycling to work is net negative or positive in terms of effect on life expectancy (breathing in more particles vs. better cardiovascular health), it seems pretty clear form this that getting an air purifier is a smart investment..
If you don’t want to read this long, looooooong article (I’m sorry) just do these things in this order:
If you have an ultrasonic humidifier, kill it.
Monitor local air quality like the weather.
No incense.
Extinguish candles with a lid.
Be careful about smoke when cooking.
Get a particle counter.
Use an air purifier at home all the time. (Move this to #1 if the outdoor air has high particulate levels where you live.)
Install a HEPA cabin air filter in your car.
Avoid aerosols.
Use a mask very carefully when in dirty air.
A strange list, admittedly, but it’s where the evidence seems to lead.
🤪 Weird properties of large language models
These days everyone is praising all the amazing capabilities of LLMs and the opportunities arising from them. I think its kind of a brave move to write an article about why LLMs are not soooo great after all. Although it very much depends on your use case whether the properties pointed out here are in fact weaknesses (or maybe features), I think its an interesting / refreshing take to think about some of the weird properties of LLMs.
Large language models (LLMs) like ChatGPT produce text that sounds very human-like. It’s easy to anthropomorphize such systems. However, the way they work internally appears to be quite different from how the human brain works.
Many weird features of LLMs have been discussed, but I’ve never seen a bunch all described in one place. This post attempts to do that. By taking stock of all of these oddities together, I believe some of fundamental limitations of LLMs start to become a bit clearer.
For a variety of reasons that I hope to articulate in upcoming posts, I am not worried about scaled-up LLMs posing any sort of existential risk. This is the first post in that series. I am worried however about LLMs being used for nefarious purposes, like spreading disinformation or helping design biological weapons.
Food for Thought.
🤖 It has been going through the news that ‘Sam Altman wants 7 trillion’. The reason is that each GPT costs between 25x and 100x as much as the last one. How much better can language models still get?
⚖️ Seems like >35% of the weight loss caused by Novo Nordisk’s hyped weight loss drug comes from bone density and muscles (and not fat). While I think this drug certainly has a potential to positively influence the lives of many, I think we should be very cautious with prescribing it on a large scale. And its certainly not a permanent ‘fix’ to whats often referred to as the ‘obesity epidemic’.
🧐 I admit this is a bit dense and I spent quite a lot of time thoroughly reading this article - but especially if you are doing any kind of Social Science research I can nonetheless recommend investing the time to read Theory-Testing in Psychology and Physics: A Methodological Paradox. Alternatively here is my two picture summary (essentially its saying that social sciences is not really credible the way it works now and that we should be making more point/range predictions):
Random Stuff.
⏰ I might have shared this before - but I just think its an interesting chart & a good starter to reflect on your own enjoyment levels and how they match up with your actual schedule. Are there any opportunities to incorporate more of the right side activities into your life?
🍳 It seems like some people are able to smell if dishes previously had eggs on them (and some are not). There is even a word for it in Arabic - zankha. See below for the results of a dinner party trial.
🐓 Sometimes I feel like I am just running around like a headless chicken all the time. Good to hear that I am not alone.
Personal Update.
Going good so far. Really enjoying that Boston has sunshine during winter (which seems to happen almost never in Copenhagen).
Lot’s of work & many classes that I am excited about (still in the honeymoon phase of arriving here I guess…)
Turned out my bike purchase was not so bad afterall - after pumping up the tires and doing some small fixes, things are working well so far. Already did two longer trips, keeping my fingers crossed nothing will break 🤞
Went to a Celtics game. Its insane how much money people spend on basketball here. I am a sports fan and happy to pay to see good games but this is really a different scale compared to what I am used to…