We commented on this in an earlier links post, now Aella goes off on the same point: You don’t need a perfectly random sample for useful data, jfc. She’s right by the way, as is Scott Alexander. See also Eigenrobot on twitter: “the biggest problem in both statistical practice and criticism of statistical practice is braindead insistence on following form rather than consideration of whether adherence to form is sufficient to produce the desired insight or necessary to produce any insight respectively”
Bad words get more differenter over time, especially adjectives: The word for good is similar in English (“good”) German (“gut”) and Faroese (“góðan”) But the word for bad is “bad” in English, “schlecht” in German, and “illur” in Faroese. In new research, we show that this is a broader pattern that we call “valence-dependent mutation”
@selentelechia on twitter has been trying an “old 4chan iodine theory on fast food salt cravings”, which is about as crazy as it sounds, but she’s experienced some pretty good outcomes: “I’ve been doing this for two days and my hands aren’t cold anymore, nor are they taking half an hour to warm up after coming inside … I don’t feel a constant low level impulse to lie down”. Seems interesting.
“Many cell lines that are widely used for biomedical research have been overgrown by other, more aggressive cells,” begins Wikipedia’s list of contaminated cell lines. “For example, supposed thyroid lines were actually melanoma cells, supposed prostate tissue was actually bladder cancer, and supposed normal uterine cultures were actually breast cancer … Estimates based on screening of leukemia-lymphoma cell lines suggest that about 15% of these cell lines are not representative of what they are usually assumed to be. … Contaminated cell lines have been extensively used in research without knowledge of their true character. For example, most if not all research on the endothelium ECV-304 or the megakaryocyte DAMI cell lines has in reality been conducted on bladder carcinoma and erythroleukemia cells, respectively.” (h/t @MasterTimBlais)
Chat GPT: Weirdly good at correcting OCR errors in historical texts. Good at condensing mind-numbing academic research into something you can actually read, just ask it to “rewrite this obtuse paper as a children’s book”. Ok at riddles until you give it a riddle with no right answer, in which case it confidently comes up with a completely nonsensical explanation. In this example, it can’t even count:
@goblinodds asks twitter, “do both of your eyes see the same colors or is one’s input cooler-toned than the other?”, finds that 20.9% report different temperature input from different eyes. Uh???
Missed opportunity: You could have owned CHARLES DICKENS’ PICKLE FORK, for the low low price of $6,120!
“Many researchers have conjectured that the humankind is simulated along with the rest of the physical universe – a Simulation Hypothesis. In this paper, we do not evaluate evidence for or against such claim, but instead ask a computer science question, namely: Can we hack the simulation?” Science Banana draws particular attention to Table 1:
Doing science online – A view on science blogging from back in 2009. And from the same author very recently: Exploratory notes: Community as the unit of scientific contribution
“I think of the spider whom, sitting like the iris inside a lacy eye, tugs and flexes and tightens its grip on different strings, creating an interrogative experience with web and with world. Scientists have likened this behavior to the activity of a brain itself, sifting through and reacting to stimuli. Each tug is a question, each returning vibration a reply. … extended cognition researcher Hilton Japyassú has shown that cutting a part of the silk dramatically shifted and disoriented the behavior of the spider, and seemed to imitate the effects of a lobotomy. This begs the question. Where is the spider’s mind? Is it inside the spider’s actual brain? Is it in its spinnerets or legs? Is it in the web itself?”
Great bloggers are rare, weird, and not team players. Showing our biases here, but we actually think that this is an argument for teams of bloggers, like yours trulies. For one person to be a great blogger they may indeed need to be obsessed about writing all the time & very widely read & interested in just about everything & willing to work for relatively low wages, but if your blogging team is two people, you only need to have that combination of traits between the two of you. If you can make a blogging collective of four people, you only need one person who has each of those traits! Maybe it’s a crazy scheme but we’re the ones with the hive mind over here.
If you ask ChatGPT to behave like a Linux terminal and start feeding it Linux commands, it will invent an entire fictional machine, complete with an entirely hallucinated internet that exists only inside ChatGPT’s language model. If you look in the letters folder, you can (sometimes) find John Doe’s resume.