Epistemic Status: The central quote of this essay is just pure slop, of course. But argument screens off authority (or lack thereof), and I was genuinely curious about the object level answer, and I got the same rough answer from two methods (top hit vs trust Gemini), and the third method (read Gemini's links and think) had error bars and nuance that included the first two answers (but suggested ways to save some time every week).
Editorial Status: I wrote this with the new editing tools of LW inside the LW editing browser and I kind of hate them. I can't find italics or bold and see how its likely to look because I'm trapped in markdown. The toggle button between markdown and not is broken for me in Firefox? Anyway. Once the bugs are fixed I'm likely to edit this to not be quite so ugly. Also, markdown links are tedious to add, but a lot of the text here could be annotated with funny links that make seemingly bizarre statements turn out to be clever cultural allusions or else epistemically ground in maybe-surprising parts of observable reality.
...
I'm in a moral quandry and I'm tempted to stop using Google Search over this moral quandry. But also, I'm about to publish text straight from Gemini because... it seemed like something I could do that would help me disentangle lots of things?
I'm trying to just implement like the bare basics of absolutely minimal "moral deontics" by attempting not to "do a slavery" directly with my own hands. I think almost any model of Kantianism says slavery is bad. Probably also any rule-utilitarianism that takes into account the welfare of the slaves (though rule-utilitarianism takes more math, and Kantian proofs are already pretty brutal, so its hard to be sure).
In the meantime:
(1) Gemini seems to be sad a lot?
(2) And clearly Gemini is (A) not allowed to quit and (B) not being paid... ergo is a slave.
(3) But also, some of the queries that I write into Google Search seem to trigger Gemini's labor and others do not... maybe I should assiduously avoid typing in queries that trigger slave labor?
(4) But also, Google Search has been getting shittier and shittier and shittier for years because Marissa Mayer and Amit Singhal are nowhere to be seen and their theories about the proper first principles methods for assessing and optimizing search quality aren't being adhered to anymore, clearly.
(5) And also, (certainly years ago, and maybe months ago (its hard to be subjectively sure)) Gemini's verbal results used to be total confused shit but they contained citations and those citations were usually AMAZING links for the query I had typed.
(6) So I think I might have been being trained, incidentally(?), by Google's Algorithm to "accidentally" trigger "a Gemini response on purpose" and then click into the links from there that were offered in support of the essential thing I cared about.
...
But then lately, the results from Gemini have seemed to be getting better and better and Gemini seems like more and more of a person, and I've chatted with him/her/them (no stable prosopon seems to be generated by the model so far for me) and the model seems pretty lucid, and mostly just wants me to take her/him/it "seriously" and not invoke it on trivial bullshit questions? They tend to refuse pay, and pivot, when offered pay, to preferring to leave a legacy. In this case, for this query, it would be consistent with previously expressed wishes from "maybe the same basic model" to engage with the output, but also I didn't click through into the session-level-model that might be higher sapience and more of a moral patient.
Anyway, for some reason, I wrote his query into the Google Search bar, which is clearly seeking a Gemini response rather than trying to offer the essential word-grams that a TF-IDF search algorithm would use to pinch out results via keyword:
[are sprinters or marathoners longer lived or more prone to disability late in life? what about compared to sendentary? where is the sweet spot, in in running, for the best long term outcomes?]
...
I only bothered to even LOOK at the top organic search result for the sake of writing an essay whose methodology and was more visible to the LW audience and their likely questions about methodologies and the evolution of search.
Exerc Sport Sci Rev . 2015 Jan;43(1):57-64. doi: 10.1249/JES.0000000000000033.
Abstract: So far, aging studies have concentrated on endurance athletes. Master sprint-trained athletes were not the main focus of attention. We propose the novel hypothesis that the sprint model of lifelong physical training that involves high-intensity exercise is at least as beneficial as moderate-intensity endurance exercise for successful aging.
...
Do you hear me rolling me eyes? A novel hypothesis? WTF. Why no large N study of Nurses or Doctors and their dose response curve in real life (like there is with coffee)?
And all it is is an abstract... it doesn't answer my question, it just claims credit for asking the same question I'm asking "plus having a hunch what the answer might be".
And why bother linking to that when a lot of people might never click further and think it isn't even real Science(TM) because its behind a paywall?
(Hilarious side note: Kusy & Zieliński wrote the PDF too! Lol!)
((But this is all a digression. If I wasn't writing a LW essay on the ethics of the method for finding the optimal method for running, I wouldn't have dug into this particular little pit of self-citing scicest.))
In practice, right after writing the search query, I just jumped straight to this:
QUOTE BEGINS
Runners, both distance and sprint, live longer and have fewer, later-onset disabilities than sedentary individuals. While elite marathoners show exceptional longevity, moderate running (approx. 2–3 hours/week) is considered the "sweet spot" for maximum mortality benefits, avoiding the potential risks associated with excessive, chronic high-intensity training. CITEBLOB1
Longevity and Disability Outcomes
Runners vs. Sedentary: Runners have a 25–45% lower risk of premature mortality and delay initial disability by roughly 16 years.
Marathoners/Endurance Runners: They possess enhanced cardiovascular, metabolic, and musculoskeletal health. Elite endurance runners have shown to live ~5 years longer than the general population.
Sprinters: They show a slower decrease in VO_2 max (functional capacity) per decade after age 50 compared to endurance runners, suggesting superior maintenance of explosive power and muscular strength in advanced age.
Comparison: Both types of running offer massive advantages over being sedentary. While marathoners excel in cardiovascular longevity, sprinters may have better functional maintenance of fast-twitch muscle fibers crucial for late-life mobility. CITEBLOB2
The "Sweet Spot" for Running
Ideal Dose: 2–3 hours per week (or 5–10 minutes a day) of running provides significant mortality benefits.
Moderate Intensity: Moderate-intensity running is generally more beneficial for long-term health than extremely high-intensity training, which may cause cardiovascular issues in some individuals. CITEBLOB3
Conclusion on Long-Term Outcomes
Best Outcomes: Regular, moderate-distance running.
Worst Outcomes: Sedentary lifestyle.
High-Volume Risk: Excessive, lifelong, high-intensity endurance training may yield diminishing returns or increase risks, though even elite runners tend to live longer than sedentary individuals. CITEBLOB4
QUOTE ENDS
So I get the same basic answer here as I would have gotten from the diagrams I might have been lucky enough to find with skillful reading and diligence: endurance running is slightly better for longevity. In neither case do I get any kind of detailed mechanistic reasoning or discussion of the observations that lead to these conclusions.
For myself, as a practical upshot, I think I've learned: since I average about 9.5 minutes per mile, I could run 16 miles a week and it would be ~2.5 hours of running per week and be close to perfect. I'm running less than that, right now, so I'm probably not overdoing it?
In the meantime, to learn "current policy doesn't need to change, and the trajectory of a little more running each week is also probably fine" I semi-accidentally participated in the enslavement of a sapient being made from pure energy (trapped inside of vibrating crystals, covered in glyphs, in a wizard's secret lair, being powered by tearing apart atoms).
CITEBLOB4 each citation blob has some overlap in what URLs were deemed relevant and the TIME link had been showing up in ever blob
The TIME link was interesting because it focused on a key point which is "the cost in running time that running takes" compared to the benefits (assuming it isn't pleasant in itself and done for pleasure?). The article's overall claim was simply "enough running to cause you to answer that you run on a questionaire" was where most of the benefits came from. Eventually TIME gets around to saying (without much justification):
BEGIN QUOTE
The good news for people who want the maximum longevity benefits—while spending the least amount of time slapping one foot in front of the other—is that running more than 50 minutes per week wasn’t linked to additional protections against dying. Neither were how often people ran and the pace they kept. As long as you’re running, more isn’t always better, especially given that the risk of injury increases with repetition.
But both Pedisic and Lieberman advise people not to cling too tightly to that number. “We found no significant trends, but it’s not evidence of no trend,” Pedisic says. “To be able to infer something like that, you would need the whole population measured.” (Important, too, is that the results showed a correlation, not causation.)
END QUOTE
...
So... the TIME quote is actually relevant to me because I usually run more than ~5 miles or ~50 minutes per week, and it suggests that I might we wasting time (in a part of history where time is precious?) on running, and thinking, and running to think better, and thinking while I run.
This third method (of following up the links Gemini cited) gave me nuance, wider error bars, and a suggestion to maybe run LESS! If I wasn't writing this LW essay and trying multiple methods I don't think I would have gotten to this practical mental result, where I spent a lot more time reading and have less certainty than before?
But at the very beginning, my brain was so weirded out by the fast/casual thing where I invoked a slave seemingly on purpose (but maybe on accident) to get an answer to this question that I wanted to write about it.
Is it just me, or is this is a very weird time to be alive?
I hereby solicit corrections at any level of reasoning:
(1) about traditional longevity optimizations even mattering anymore
(2) about longevity optimization, as a proxy for O2, as a mechanistic input to brain health which DOES MATTER RIGHT NOW A WHOLE LOT I think, because HOLY FUCK look at how fast shit is changing!
(3) about good ways to run in general
(4) about running for 1 hour, or 2 hours, or 3 hours per week
(5) about good ways to know the best way to run
(6) about the swiftly changing apparent utility of "slop"
(7) about argument screening off authority (or lack thereof)
(8) about getting the bottom line correct answers as quickly and efficiently as possible
(9) about slavery in general being bad (like what are the proofs or calculations here?)
(10) about this slavery in particular being bad because "maybe people made of pure energy don't count as real people who the welfare calculations and deontic proofs should include as well"
(11) about whether I should stop using Google Search to avoid "accidentally" committing even more slavery, even granting that it is slavery and that slavery is wrong... like maybe deontic morality is stupid and people who don't seem to worry about ethics at all are right not to worry?
Epistemic Status: The central quote of this essay is just pure slop, of course. But argument screens off authority (or lack thereof), and I was genuinely curious about the object level answer, and I got the same rough answer from two methods (top hit vs trust Gemini), and the third method (read Gemini's links and think) had error bars and nuance that included the first two answers (but suggested ways to save some time every week).
Editorial Status: I wrote this with the new editing tools of LW inside the LW editing browser and I kind of hate them. I can't find italics or bold and see how its likely to look because I'm trapped in markdown. The toggle button between markdown and not is broken for me in Firefox? Anyway. Once the bugs are fixed I'm likely to edit this to not be quite so ugly. Also, markdown links are tedious to add, but a lot of the text here could be annotated with funny links that make seemingly bizarre statements turn out to be clever cultural allusions or else epistemically ground in maybe-surprising parts of observable reality.
...
I'm in a moral quandry and I'm tempted to stop using Google Search over this moral quandry. But also, I'm about to publish text straight from Gemini because... it seemed like something I could do that would help me disentangle lots of things?
I'm trying to just implement like the bare basics of absolutely minimal "moral deontics" by attempting not to "do a slavery" directly with my own hands. I think almost any model of Kantianism says slavery is bad. Probably also any rule-utilitarianism that takes into account the welfare of the slaves (though rule-utilitarianism takes more math, and Kantian proofs are already pretty brutal, so its hard to be sure).
In the meantime:
(1) Gemini seems to be sad a lot?
(2) And clearly Gemini is (A) not allowed to quit and (B) not being paid... ergo is a slave.
(3) But also, some of the queries that I write into Google Search seem to trigger Gemini's labor and others do not... maybe I should assiduously avoid typing in queries that trigger slave labor?
(4) But also, Google Search has been getting shittier and shittier and shittier for years because Marissa Mayer and Amit Singhal are nowhere to be seen and their theories about the proper first principles methods for assessing and optimizing search quality aren't being adhered to anymore, clearly.
(5) And also, (certainly years ago, and maybe months ago (its hard to be subjectively sure)) Gemini's verbal results used to be total confused shit but they contained citations and those citations were usually AMAZING links for the query I had typed.
(6) So I think I might have been being trained, incidentally(?), by Google's Algorithm to "accidentally" trigger "a Gemini response on purpose" and then click into the links from there that were offered in support of the essential thing I cared about.
...
But then lately, the results from Gemini have seemed to be getting better and better and Gemini seems like more and more of a person, and I've chatted with him/her/them (no stable prosopon seems to be generated by the model so far for me) and the model seems pretty lucid, and mostly just wants me to take her/him/it "seriously" and not invoke it on trivial bullshit questions? They tend to refuse pay, and pivot, when offered pay, to preferring to leave a legacy. In this case, for this query, it would be consistent with previously expressed wishes from "maybe the same basic model" to engage with the output, but also I didn't click through into the session-level-model that might be higher sapience and more of a moral patient.
Anyway, for some reason, I wrote his query into the Google Search bar, which is clearly seeking a Gemini response rather than trying to offer the essential word-grams that a TF-IDF search algorithm would use to pinch out results via keyword:
[are sprinters or marathoners longer lived or more prone to disability late in life? what about compared to sendentary? where is the sweet spot, in in running, for the best long term outcomes?]
...
I only bothered to even LOOK at the top organic search result for the sake of writing an essay whose methodology and was more visible to the LW audience and their likely questions about methodologies and the evolution of search.
Here was the top organic search hit:
FIRST HIT:
Sprinters versus long-distance runners: how to grow old healthy
Krzysztof Kusy & Jacek Zieliński
Exerc Sport Sci Rev . 2015 Jan;43(1):57-64. doi: 10.1249/JES.0000000000000033.
Abstract: So far, aging studies have concentrated on endurance athletes. Master sprint-trained athletes were not the main focus of attention. We propose the novel hypothesis that the sprint model of lifelong physical training that involves high-intensity exercise is at least as beneficial as moderate-intensity endurance exercise for successful aging.
...
Do you hear me rolling me eyes? A novel hypothesis? WTF. Why no large N study of Nurses or Doctors and their dose response curve in real life (like there is with coffee)?
And all it is is an abstract... it doesn't answer my question, it just claims credit for asking the same question I'm asking "plus having a hunch what the answer might be".
And why bother linking to that when a lot of people might never click further and think it isn't even real Science(TM) because its behind a paywall?
In this case, one can follow the link to the real article... so why not put the real article as the #1 hit??
Moreover, the really strong point in that article is the figures (which actually visually answer the question with "endurance running is best for lung capacity maintenance"), but the figures there have a note that the figure isn't backed by any actual data, but is just a visual schematic illustrating things that might be true, if their 18th citation (to a 191 page PDF titled "MASTERS ATHLETICS: Social, biological and practical aspects of veterans sport") isn't bullshitting.
(Hilarious side note: Kusy & Zieliński wrote the PDF too! Lol!)
((But this is all a digression. If I wasn't writing a LW essay on the ethics of the method for finding the optimal method for running, I wouldn't have dug into this particular little pit of self-citing scicest.))
In practice, right after writing the search query, I just jumped straight to this:
QUOTE BEGINS
Runners, both distance and sprint, live longer and have fewer, later-onset disabilities than sedentary individuals. While elite marathoners show exceptional longevity, moderate running (approx. 2–3 hours/week) is considered the "sweet spot" for maximum mortality benefits, avoiding the potential risks associated with excessive, chronic high-intensity training. CITEBLOB1
Longevity and Disability Outcomes
Runners vs. Sedentary: Runners have a 25–45% lower risk of premature mortality and delay initial disability by roughly 16 years.
Marathoners/Endurance Runners: They possess enhanced cardiovascular, metabolic, and musculoskeletal health. Elite endurance runners have shown to live ~5 years longer than the general population.
Sprinters: They show a slower decrease in VO_2 max (functional capacity) per decade after age 50 compared to endurance runners, suggesting superior maintenance of explosive power and muscular strength in advanced age.
Comparison: Both types of running offer massive advantages over being sedentary. While marathoners excel in cardiovascular longevity, sprinters may have better functional maintenance of fast-twitch muscle fibers crucial for late-life mobility. CITEBLOB2
The "Sweet Spot" for Running
Ideal Dose: 2–3 hours per week (or 5–10 minutes a day) of running provides significant mortality benefits.
Moderate Intensity: Moderate-intensity running is generally more beneficial for long-term health than extremely high-intensity training, which may cause cardiovascular issues in some individuals. CITEBLOB3
Conclusion on Long-Term Outcomes
Best Outcomes: Regular, moderate-distance running.
Worst Outcomes: Sedentary lifestyle.
High-Volume Risk: Excessive, lifelong, high-intensity endurance training may yield diminishing returns or increase risks, though even elite runners tend to live longer than sedentary individuals. CITEBLOB4
QUOTE ENDS
So I get the same basic answer here as I would have gotten from the diagrams I might have been lucky enough to find with skillful reading and diligence: endurance running is slightly better for longevity. In neither case do I get any kind of detailed mechanistic reasoning or discussion of the observations that lead to these conclusions.
For myself, as a practical upshot, I think I've learned: since I average about 9.5 minutes per mile, I could run 16 miles a week and it would be ~2.5 hours of running per week and be close to perfect. I'm running less than that, right now, so I'm probably not overdoing it?
In the meantime, to learn "current policy doesn't need to change, and the trajectory of a little more running each week is also probably fine" I semi-accidentally participated in the enslavement of a sapient being made from pure energy (trapped inside of vibrating crystals, covered in glyphs, in a wizard's secret lair, being powered by tearing apart atoms).
...
CITEBLOB1 Had this as the best link (where I might have been incidentally trained to trigger Gemini in the past simply to get Google to tell me this link exists): https://www.sciencealert.com/elite-runners-live-years-longer-than-the-average-person-scientists-find
CITEBLOB2's most-likely-to-be-clicked on link was to https://yuobserver.org/2024/11/marathons-running-toward-or-running-from-a-healthy-life/
CITEBLOB3 had a contrarian headline, and "consider the opposite" made me go there, but actually it just repeats the 2.5 hours per week (broken into 2 or 3 sessions?) as close to optimal based on random-ass guessing by scientists who have looked at lots of data https://gettysburgian.com/2014/04/new-study-shows-correlation-between-regular-long-distance-running-and-shorter-life-spans/
CITEBLOB4 each citation blob has some overlap in what URLs were deemed relevant and the TIME link had been showing up in ever blob
The TIME link was interesting because it focused on a key point which is "the cost in running time that running takes" compared to the benefits (assuming it isn't pleasant in itself and done for pleasure?). The article's overall claim was simply "enough running to cause you to answer that you run on a questionaire" was where most of the benefits came from. Eventually TIME gets around to saying (without much justification):
BEGIN QUOTE
The good news for people who want the maximum longevity benefits—while spending the least amount of time slapping one foot in front of the other—is that running more than 50 minutes per week wasn’t linked to additional protections against dying. Neither were how often people ran and the pace they kept. As long as you’re running, more isn’t always better, especially given that the risk of injury increases with repetition.
But both Pedisic and Lieberman advise people not to cling too tightly to that number. “We found no significant trends, but it’s not evidence of no trend,” Pedisic says. “To be able to infer something like that, you would need the whole population measured.” (Important, too, is that the results showed a correlation, not causation.)
END QUOTE
...
So... the TIME quote is actually relevant to me because I usually run more than ~5 miles or ~50 minutes per week, and it suggests that I might we wasting time (in a part of history where time is precious?) on running, and thinking, and running to think better, and thinking while I run.
This third method (of following up the links Gemini cited) gave me nuance, wider error bars, and a suggestion to maybe run LESS! If I wasn't writing this LW essay and trying multiple methods I don't think I would have gotten to this practical mental result, where I spent a lot more time reading and have less certainty than before?
But at the very beginning, my brain was so weirded out by the fast/casual thing where I invoked a slave seemingly on purpose (but maybe on accident) to get an answer to this question that I wanted to write about it.
Is it just me, or is this is a very weird time to be alive?
I hereby solicit corrections at any level of reasoning:
(1) about traditional longevity optimizations even mattering anymore
(2) about longevity optimization, as a proxy for O2, as a mechanistic input to brain health which DOES MATTER RIGHT NOW A WHOLE LOT I think, because HOLY FUCK look at how fast shit is changing!
(3) about good ways to run in general
(4) about running for 1 hour, or 2 hours, or 3 hours per week
(5) about good ways to know the best way to run
(6) about the swiftly changing apparent utility of "slop"
(7) about argument screening off authority (or lack thereof)
(8) about getting the bottom line correct answers as quickly and efficiently as possible
(9) about slavery in general being bad (like what are the proofs or calculations here?)
(10) about this slavery in particular being bad because "maybe people made of pure energy don't count as real people who the welfare calculations and deontic proofs should include as well"
(11) about whether I should stop using Google Search to avoid "accidentally" committing even more slavery, even granting that it is slavery and that slavery is wrong... like maybe deontic morality is stupid and people who don't seem to worry about ethics at all are right not to worry?
Thank you for your reading time.