I think I just discovered a new weird cultural linguistic quirk of American English. If you go to a grocery store and ask “where are the peaches?”, you will get directed to the fresh produce section. There are probably also peaches in the canned fruit section and in the frozen fruit area, but by default “peaches” is understood to refer to fresh whole peaches. When looking for peaches, we typically use labels “peaches” (for fresh), “canned peaches” and “frozen peaches”. Similarly, if you ask “where is the chicken?” you get sent to the butcher area where there are refrigerated whole chickens. American English speakers do not consider frozen chicken nuggets, deli-sliced chicken, or cooked rotisserie chicken to be the default “chicken”. In general, if you ask the location of a food item in the grocery store, the default is to assume you are referring to the fresh/unprocessed/uncooked version of the item. The exception is bread. If you ask “where is the bread”, you do not get sent to the fresh bread in the bakery section, but to the bread aisle where the sliced shelf stable bread is. Rather than follow the rules of other food items, we have “bread” and “fresh bread” rather than “bread” and “sliced bread”. If an American English speaker wants bread from the bakery, we use terms like “French bread”, “baguette”, or even “crusty bread”.
I think this really pisses off Germans, since they seem to have the opposite default where das Brot is understood to be fresh sliced bread and Toastbrot is pre-sliced industrial bread.
It’s interesting that Harvard says “Homicide leading cause of death”, while the other headlines have “homicide a leading cause of death” (emph added on that a). I think that changes the meaning of the sentence. Without the “a”, I think the missing article most US English speakers would assume is “the”.
It looks like these news articles all trace back to an opinion (i.e., not peer reviewed) piece in the British Medical Journal, which in turn cites an Obst Gynecol articleHomicide During Pregnancy and the Postpartum Period in the United States, 2018-2019. Now that article has the sentence “Homicide during pregnancy or within 42 days of the end of pregnancy exceeded all the leading causes of maternal mortality by more than twofold”. But if we look at their plot, They can only say that because they split all the different medical causes of death into different categories! They don’t have a table with the numbers that go into the plot, but if one is petty enough to load it into an image viewer and count the damn pixels, one finds that the medical complications actually outnumber homicides.
The paper also states they find “3.62 homicides per 100,000 live births among females who were pregnant or within 1 year postpartum”. Again, that’s bad. There should be zero homicides per 100,000 births. Homicides are very bad. But, just to put that number in perspective, the US is currently has around 12 motor vehicle deaths per 100k people. And the homicide rate for men is 5.4 per 100k.
This is tough, we’ve got Harvard on one side saying gun violence is “a health emergency for pregnant women.”. But when we look at the data around how people in the US die, it looks like homicide accounts for around 10% of pregnant women deaths, I would call it probably the 4th leading cause of death behind drug overdoses, auto accidents, and medical complications. This NEMJ article Overdose, Homicide, and Suicide as Causes of Maternal Death in the United States seems to track that.
How does one go against fighting a clickbait social panic? I’m going to call this the Nuclear Winter Problem. In the early 1980s, everyone’s favorite astronomer Carl Sagan spoke out against nuclear weapons. Sagen published research about how the resulting fire storms in cities hit with nuclear weapons would pollute the atmosphere and cool the planet, causing a “nuclear winter”. Scientists in the field found his work to be lacking, he seemed to me massively over-predicting the cooling effect. But no one wants to be seen as speaking out in favor of nuclear weapons.
Random ideas that I don’t have time to follow up on
I should write a twitter bot that finds and retweets jokes that are variations of “teach a man to fish…”. It would be a good ML natural language kind of project
I should write a twitter bot that finds and retweets trolly problem memes. It would be a fun ML vision project
I should scrape twitter for folks posting wordle scores and see what fraction of folks are playing “wrong”, ie, playing as if they are in hard-mode, but not actually in hard-mode. It’s been eye-opening to see how many folks will lock in their correct letters even if it would clearly be better to eliminate many potential letters in a turn.
One of my favorite ways to play Jeopardy is to try and guess what some of the answers will be based only on the catagory. For example, if the catagory is “Bond Films”, I might guess Goldeneye, Live and Let Die, and Sean Connery as answers I expect to show up at some point. I want an app that lets me guess answers ahead of time and gives huge bonus points if I manage to hit one. Even bigger bonus if it’s a daily double. And if you guess the final jeopardy answer from the topic alone you win the game.
I like this paper though, turns out booster seats, car seats, and seat belts are all about the same effectiveness. Juts strap kids down some way, it doesn’t really matter. https://www.sciencedaily.com/releases/2018/01/180105124030.htm
