isopod's recent activity
-
Comment on Even with the reactionary backlash, trans acceptance has been the one good news in this millennium in ~lgbt
-
Comment on Why therapy is broken in ~health.mental
isopod (edited )LinkI think this article points out some very real issues. The current crisis did not materialize out of nothing, though, and the solutions the authors suggest - universal mental health care...- Exemplary
I think this article points out some very real issues. The current crisis did not materialize out of nothing, though, and the solutions the authors suggest - universal mental health care subsidies, algorithmic patient-therapist pairing, incentives to drive more psychotherapeutic education - all require a certain concentration of political will that I don't see materializing. It's a nice thought, in other words.
I think everyone in the field, patients alike, are well aware there are issues. But where do you start? Take the idea of better patient-therapist pairing, for instance. The goal is to pair patients with the right therapist, but how do you quantify what patients need, identify the core issues, understand how different therapists operate? If the therapeutic relationship is this nebulous art, a complex dance, how are you going to reduce it to something we can put into a database and host on a web server? There's probably some headway we could make into this problem today, but the act of helping another person is in no way trivial. Human growth, healing, fulfillment... we can nibble at the edges of these, but we're not going to capture them. That's philosophy, and we aren't there yet. We could educate patients about treatment methods and maybe encourage primary care providers to more accurately refer patients for mental health services, but I don't know - algorithmic matching feels like pie in the sky thinking.
Right now, what we've got is this: A person visits a therapist, they talk, the therapist describes a treatment approach, and the patient decides if it's right for them. If not, they bounce around a bit and find someone who fits better. Are you really gonna improve that process by inserting yet another algorithm?
I don't think the lack of evidence-based methods in therapy is necessarily a weakness. Take the studies cited in the article that grab a random sample of patients and ask them whether some therapy helped. How many of them say "yes" even though deep down they're just hopeful, they like their therapist, or they're being polite? How many say "no" because they're resisting a change they need to make or because they had some blown-up idea of what therapy could do? How easy is it for anyone to reduce their progress as a human being to a response like that? A lot of noise in that signal.
There are certainly ways we could improve mental health. Funding and education, for sure. But even if we had the political will - which I wish we did - people who are in healthy communities and homes, who have things to look forward to, and who are in a functioning society, free of fear and want, are generally a lot less likely to have severe mental health issues. And if we want to point a finger, maybe allocate some resources, I feel maybe we should consider aiming it at the fact that our world falls far short of that. The mental health profession can certainly be improved - but if we don't address the big picture, to some degree, I think we're just going to get more adept at putting band-aids on wounds.
-
Comment on Have you or anyone in your family ever won anything? in ~talk
isopod LinkNothing big, but we won a plasma TV in a raffle back when they were brand new. The thing must have weighed 40 kg. It was like a big squooshed brick. Contrary to some windfalls, we used it up to...Nothing big, but we won a plasma TV in a raffle back when they were brand new. The thing must have weighed 40 kg. It was like a big squooshed brick.
Contrary to some windfalls, we used it up to its untimely death a year or so ago. If it had been a million dollars, ehh, might have been a bit more reckless...
-
Comment on Something wrong with new game reviews (Dave the Diver) in ~games
isopod LinkOpinion is subjective. My partner LOVES the game, I mean, can't put it down for hours, literally salivating to find out what happens next. It might not be your bag, but you might not be the same...Opinion is subjective. My partner LOVES the game, I mean, can't put it down for hours, literally salivating to find out what happens next. It might not be your bag, but you might not be the same as the next person.
If you want to point fingers at something, I'd point them at the very idea that a single number can capture how good a game is, overall, for "all of us". If you think that, either as a reviewer or a reader, I think you're making a mistake. The best reviews are those where the reviewer's perspective and bias are a feature rather than a bug.
-
Comment on What happened in 1971? (various graphs of U.S. economic data, showing turning points in or around 1971) in ~finance
isopod Link ParentHey! I was hoping my comment would inspire discussion that I could learn from, so thank you for taking the time to reply. I'm only an interested outsider in finance. Let me know if I'm...Hey! I was hoping my comment would inspire discussion that I could learn from, so thank you for taking the time to reply. I'm only an interested outsider in finance.
Let me know if I'm understanding your points:
- The transition to fiat currency led to a period of unprecedented stability, not instability;
- The gold standard may not have been a fiat system in name, but it was a fiat system in practice;
- Inflation is good in moderation.
Looking over your comment and mine, I'm wondering if we're just using different semantics. The instability I was thinking of was just the fact that a lot of the graphs in OP's link start trending differently in the 1970's than they had before. From the perspective of the international financial system's stability, I'm quite sure you are correct. I also think you're right about the gold system, which I always thought was kind of silly on the face of it (what inherent value does gold have, anyway?), but I'm wondering if there are other dimensions to Nixon's decision or its consequences that might have left echoes beyond that.
In short, in the absence of a better explanation, I was thinking: What happened in 1971, indeed? Perhaps there's something to this? Bretton Woods was the first thing that came to mind.
The one thing I'm concerned about is the idea that
if you work your wages will grow to keep up with the cost of living (albeit at a lag)
Does the data bear that out in practice? I'm seeing the opposite trend in the graphs. Maybe I'm missing something.
-
Comment on GPT detectors are biased against non-native English writers in ~tech
isopod Link ParentOh, no, I literally asked GPT-4 to take its own output and rewrite it in a different style with more surprising word choice -- literally, to "increase the perplexity". I didn't save the entire...Oh, no, I literally asked GPT-4 to take its own output and rewrite it in a different style with more surprising word choice -- literally, to "increase the perplexity". I didn't save the entire prompt but it generally comes out something like:
Please rewrite the above passage to use more surprising, varied, and interesting word choice, employing poetic and interesting diction, and in general, increase its perplexity.
GPT-4 is quite capable at following instructions like this. You might have to tweak the prompt for your particular use case, of course.
-
Comment on What happened in 1971? (various graphs of U.S. economic data, showing turning points in or around 1971) in ~finance
isopod LinkComplete speculation, but... One possibility is that 1971 was when Nixon effectively killed the Bretton Woods system, an international agreement that tied currency values to the US dollar, which...Complete speculation, but...
One possibility is that 1971 was when Nixon effectively killed the Bretton Woods system, an international agreement that tied currency values to the US dollar, which was in turn tied to gold. This led countries to transition to fiat currency, which basically means free-floating money not backed by any physical stuff. In turn, this gave countries more flexibility in their monetary policy, but also led to increased inflation and economic instability. (Cynically speaking, this also might have helped the rich, who had the knowledge and means to profit from these changes...) So, in conjunction with the social changes happening domestically in the US, maybe that might explain why 1971 showed such a visible change?
Of course, the fact that this change has continued unabated for 50 years is probably a lot more complex than that. It's just one factor that might have been part of it.
-
Comment on The controversial gay priest who brought vigilante justice to San Francisco's streets in ~lgbt
isopod (edited )LinkIt's interesting to see how protest movements, at least when taken in retrospect, seem to get presented as either "peaceful and respectful" or "aggressive and violent." Like, in civil rights,...It's interesting to see how protest movements, at least when taken in retrospect, seem to get presented as either "peaceful and respectful" or "aggressive and violent."
Like, in civil rights, you've got Martin Luther King versus Malcolm X. In gay rights, you've got the Lavender Panthers versus the mostly white, middle-to-upper class voters that legalized gay marriage. If you consider anti-capitalism, you've got Occupy Wall Street on the one hand and quiet quitting on the other.
I'm not a fan of violence or accommodation: I'm a fan of what works. And I suspect this author is too. You probably need to have a bit of both worlds to make a healthy and effective counterculture. But whenever I read sources like this, the authors always point out, somewhere, the contrast between the two perspectives. It's just a passing thought, but I wonder if the similarities might be even more compelling.
-
Comment on GPT detectors are biased against non-native English writers in ~tech
isopod (edited )LinkA little background for interested folks... Perplexity It's likely that most GPT detectors are using perplexity as their secret sauce algorithm. Perplexity is a measurement of how "surprising" or...- Exemplary
A little background for interested folks...
Perplexity
It's likely that most GPT detectors are using perplexity as their secret sauce algorithm. Perplexity is a measurement of how "surprising" or "unexpected" the words in a text are from the perspective of average human writing.
For instance, if I said,
The man wore a kilt and played the [...]
GPT-3 predicts that the next "token", or chunk of text, is "bag" with a 97.5% probability. And of course it's leading to bagpipes. But if I were a human, I might want to be funny or surprise you, and I might say something like
The man wore a kilt and played the saxophone
In this case, GPT-3 tells me that the token "sax" has a probability of 0.0000406%. This is a very, very unlikely continuation from a statistical perspective - but not unusual for a human!
We humans differ from each other (and thus from the norm). We also intentionally surprise, vary our style, or use humor. Thus, human speech and writing have much higher perplexity than LLM outputs, which much more often stick to "likely" outputs.
And that's the point of the paper quoted by OP: Foreign language learners have less diverse patterns of speech, smaller vocabularies, and more restricted grammatical patterns, which decreases the perplexity of their speech and thus brings them into the range of what these detection systems would classify as AI-generated.
Circumvention
Incidentally, this why asking ChatGPT to rewrite essays in a different style increases their perplexity: By giving GPT a goal, it chooses words from a probability distribution shaped by that intent. For instance, the AI-generated blurb below:
Venice, known as the floating city, is a mesmerizing Italian gem built on a lagoon. Its captivating landscape boasts charming canals winding through stunning architecture, making it a haven for romantics and history enthusiasts alike.
has an average probability of -2.18 logprob/token (more negative is higher perplexity), but when I ask GPT-4 to rewrite to increase perplexity, it gives
Dancing on top of gentle waves, Venice is a whispering watercolor weaving a tale of Italian wonder. With rippling canals playing hide and seek between vibrant buildings, it's a labyrinth of love and history, studding hearts with an unfamiliar yet enchanting riddle.
which gives -3.05 logprob/token (much more surprising and human). In comparison, I wrote this myself:
Who could forget Venice, that quaint jewel of the Mediterranean? Take a gondola down its winding waterways, visit a theatre, eat lunch at the Piazza San Marco: The possibilities are endless.
This scores -2.86 logprob/token. And I'm trying not to look like an AI. I can't imagine how hard it is for folks who are second language learners.
It is not difficult to defeat AI detection if you know what to look for. Perhaps the ensuing arms race will change this.
-
Comment on What consumeristic and somewhat pointless hobby do you have? in ~hobbies
isopod Link ParentYeah, I've always struggled with the question of what to do with finished Lego builds. They're not small, I live in an apartment, and there's only so much shelf space… Do I take it apart? Dump it...Yeah, I've always struggled with the question of what to do with finished Lego builds. They're not small, I live in an apartment, and there's only so much shelf space… Do I take it apart? Dump it in the ocean and contribute to the Great Pacific Garbage Patch? Set it aflame and send it to the sky gods? Donate them to a thrift store?
-
Comment on There’s finally a psychedelic caucus in congress — here’s what they’re doing in ~health
isopod LinkI'm hopeful that state and federal governments will liberalize drug enforcement, but I'm not holding my breath. If anyone in the US is interested in their state's psychedelics law, this website...I'm hopeful that state and federal governments will liberalize drug enforcement, but I'm not holding my breath.
If anyone in the US is interested in their state's psychedelics law, this website has a very detailed breakdown. It also has links to the federal and state laws that are under consideration or that have changed recently.
As an aside, OP, that link has the most hallucinogenic design I think I've ever seen. Maybe it only gives headaches to sober people. That could be the problem...
-
Comment on Smoke will keep pouring into the US as long as fires are burning in Canada. Here’s why they aren’t being put out. in ~enviro
isopod LinkInterestingly, the quantity of fire starts hasn't really changed much this year versus the average. It's just a bit higher. However, the area burned is much greater, indicating that the fires that...Interestingly, the quantity of fire starts hasn't really changed much this year versus the average. It's just a bit higher. However, the area burned is much greater, indicating that the fires that do start are consuming much more land before petering out.
The official statistics page has all these graphs and more. It's quite interesting. Scroll down to the bottom to see the graph of area burned per year.
-
Comment on Flattening ASTs and other compiler data structures in ~comp
isopod LinkAs I was reading, I was thinking, "Why not just use bytecode?" Bytecode would be contiguous in memory, compact, and optimizable during generation. So, basically most of the benefits that...As I was reading, I was thinking, "Why not just use bytecode?" Bytecode would be contiguous in memory, compact, and optimizable during generation. So, basically most of the benefits that flattening an AST into an array of nodes would give.
Then the author arrives at the same conclusion at the end. Alas!
Nonetheless, I think that this article makes an interesting point that in simple projects or constrained environments, the approach of storing AST nodes in an array might be a pretty simple way to increase efficiency and locality. Bytecode is a lot of work, and that approach is dead simple. It's an interesting take.
-
Comment on The Titan submersible was “an accident waiting to happen” in ~transport
isopod LinkAs Richard Feynman said, Feels like Stockton Rush sorely needed to take that advice. There's no glamour in taking stupid chances, breaking rules, and being an "adventurer", if it is contrary to...As Richard Feynman said,
The first principle is that you must not fool yourself and you are the easiest person to fool.
Feels like Stockton Rush sorely needed to take that advice. There's no glamour in taking stupid chances, breaking rules, and being an "adventurer", if it is contrary to reality. And apparently, OceanGate used unproved materials, misinterpreted safety factors, dodged operational and regulatory expenses, refused to consult with experts, and threatened legal action against whistleblowers. Under those circumstances, it's no surprise that disaster soon followed.
I wonder what changes in the legal code would prevent this kind of tragedy in the future. Given the geopolitical reality, I don't honestly think it's possible. But maybe the widespread media coverage will give people pause until a few generations cycle through and the next round of foolhardy explorers forget the past.
-
Comment on The young miners dying of “an old man’s disease” in ~life
isopod LinkLike many of you, this is a story I've heard before, but here's the kicker that made me share it: Black lung is about 3x more prevalent amongst coal miners today than 20 years ago. Witness the...Like many of you, this is a story I've heard before, but here's the kicker that made me share it: Black lung is about 3x more prevalent amongst coal miners today than 20 years ago. Witness the figure. I thought we were making progress; what gives?
The resurgence of black lung appears to be from a mix of regulatory and legal changes, underfunded enforcement, lax safety standards, and cheating/fabrication by mine owners. It's sad.
I'll leave a couple quotes from the article:
“There’s no such thing as a good coal company; some are just worse than others,” according to King. “The law lets them take so many samples and then they can choose the best ones. … I remember working nonunion; when the federal government would come to run the surveys, the company would always send extra help on the bolt machines or on the scoops to get a cleaner reading than what normally they would."
MSHA has been heavily criticized for going easy on violators. Hamby found that, between 2000 and 2011, the agency received more than 53,000 samples from underground coal miners that showed overexposure to coal dust, but only about 2,400 citations were issued.
Let this be another reminder that progress is not permanent and that change contrary to human nature requires constant effort...
-
The young miners dying of “an old man’s disease”
13 votes -
Comment on Life in the cosmos: JWST hints at lower number of habitable planets in ~space
isopod LinkThe headline, as usual, plays up the research, which is primarily about establishing that a single planet, TRAPPIST-1c, has no atmosphere. But the research methodology is actually quite a bit more...- Exemplary
The headline, as usual, plays up the research, which is primarily about establishing that a single planet, TRAPPIST-1c, has no atmosphere. But the research methodology is actually quite a bit more interesting than the headline.
The journal article summarized by the link is here. I'm not a specialist, but I was interested and did some work to grasp the main ideas, which I'll try to summarize here for anyone else who wants to go deeper.
- TRAPPIST-1 is a dwarf star about 40 light years away. That's quite close in interstellar terms, which makes it easy to observe.
- TRAPPIST-1 has seven planets. The planet of interest to this study, TRAPPIST-1c, orbits its star once every 2.4 days and is about the radius of the Earth. So everything is much more compact than the solar system we live in.
- In part because TRAPPIST-1c is so close to its sun, it is tidally locked, which means that one side of it always faces its sun. (Our moon is also tidally locked to the Earth, which is why we never see its back side.)
- Therefore, the side of TRAPPIST-1c that faces its sun is expected to be much hotter than the side that faces away - unless there is some sort of atmosphere which would act to spread out the heat. This is the fulcrum that the study leans on to come to its conclusion.
Now, we can begin to unpack the abstract:
We measure a planet-to-star flux ratio of fp/f⁎ = 421 ± 94 parts per million (ppm), which corresponds to an inferred dayside brightness temperature of 380 ± 31 K. This high dayside temperature disfavours a thick, CO2-rich atmosphere on the planet.
This makes sense now: If there were atmosphere, the dayside (facing the sun) would not be so hot. But how did they figure out how hot the atmosphere is, and what does that have to do with eclipses?
The key is in the term "planet-to-star flux ratio". Flux refers to how much energy passes through a particular area (e.g. a telescope's lens) per unit time. Brightness, basically. When the planet TRAPPIST-1c passes behind its star, for a moment, its light is blocked out by the star and the brightness of the entire system dims. This figure shows a red curve which models the brightness of the system during these eclipses. Notice the red curve sits at a y-coordinate of about 400 except for the eclipse dip? That 400-ish number is the planet-to-star flux ratio. It's just a way of saying "this is how much the light dims when the planet goes behind the star".
Hotter planets give off more flux. So the amount of dimming during eclipses gives the researchers a way to infer the temperature of the planet. (And since no light comes from its dark side, we know the flux is caused by its dayside, which is the part we're trying to measure!) This is how they conclude the temperature is about 380 K.
Then they rule out the likelihood of a thick atmosphere by comparing these results to a bunch of models and simulations and showing that they do not match up with the simulations where there is an atmosphere but they are consistent with bare rock.
Whew. And that's the research. Again, not a specialist, so take all of this with a cupful of salt, but I hope it's interesting!
-
Comment on Creatures that don't conform: Slime molds and their fascinating existence in ~science
isopod LinkI can really feel the author's deep respect for the natural world. Jones writes: Maybe if we can develop empathy for lifeforms as simple and ubiquitous, yet inscrutable and diverse, as slime...I can really feel the author's deep respect for the natural world. Jones writes:
How do we see the world as sacred again? By radical noticing. Looking for awe in all of life. Following the wonder in our bodies electric. Before we find new stories, don’t we need to sit and remember? How to venerate the world?
More and more, I think a solution is awe. As Dacher Keltner’s work shows, awe seems to orient us to things outside of our individual selves. It suggests our true nature is collective. Studying narratives of awe in cultures across the world, Keltner and colleagues found that a common part of natural awe is the sense that plants and animals are conscious and aware.
Maybe if we can develop empathy for lifeforms as simple and ubiquitous, yet inscrutable and diverse, as slime molds, we can connect with anything? (Oh no: I've developed empathy for gonorrhea!)
Personally, I don't need a moral takeaway to like the slimes; they're cute enough as-is!
-
Comment on <deleted topic> in ~life
isopod LinkAround this time last year, my grandfather died at 88 years old. It was cancer; it had started in his colon and spread to his lungs. I'd lived with him since I was 18. Up until I was 36, we kept...Around this time last year, my grandfather died at 88 years old. It was cancer; it had started in his colon and spread to his lungs.
I'd lived with him since I was 18. Up until I was 36, we kept up that alliance, transitioning from surrogate father/son to something more akin to longtime roommates and finally to caretaker/caretakee as he became weaker and more dependent. I went through three relationships and got engaged all while living with him. It was a long time.
The day when he finally entered hospice and went into full-time residential care - when he left for good, never to return - was probably the single most emotional day of my life. More emotional than the moment he died, more than when I found him weeping after my grandmother sustained the injury that would kill her, more than any breakup or love I'd had. It wasn't a good emotion or a bad emotion but all of the emotions, because it was all of life, converging at a point, and then that point converging into nothing at all, and that horrible realization, that moment when the finality of it, which I knew but I did not know, was given physical form, was manifest before me. If only I could cling to that moment, prevent him from stepping out that door, maybe, somehow, I could hold off the tide...
It's been a year; I'd like to say I'm over it. And life certainly has gone on. But I'm going to have to make that journey one day, as will everyone else I know and love, so, really, there's no over to get over. That feeling is stained into me and I will carry it until the day I die, when someone else who loves me will get to shoulder the burden on my behalf. And so it goes.
Thanks, OP, for linking this very poignant post.
-
Comment on Brainfuck just-in-time compiler with hand-rolled code generation in ~comp
isopod Link ParentOh, that's very cool! I didn't realize. I think my biggest takeaway from reading your code is that if you replaced code_gen with something more powerful and wrote an AST for a slightly more...Oh, that's very cool! I didn't realize.
I think my biggest takeaway from reading your code is that if you replaced
code_genwith something more powerful and wrote an AST for a slightly more complex language, you could actually get a JIT for something nontrivial. Like, it might be a few thousand lines, but it would work.I've written a bunch of interpreters but I never actually thought it was, idk, in my capacity to do something like that. It always felt like something big teams attacked, not a side project. It's tempting, isn't it?
The most revolutionary thing I've seen as a tutor for 16-or-so years is that the standard, normally-accepted viewpoint amongst young people in my fairly liberal, east coast US area has very obviously shifted in stages:
I know it hasn't panned out this well everywhere, but as a kid of two moms who is committed to a trans partner, I'm just really glad that I can breathe easy and my partner can get proper medical care.
If the situation backslides, at least we've got a strong foothold to fight from.