I'm building a tiny experimental message board / ...game? called Isles (https://isles.app/about). The short pitch is that when you sign up you're randomly assigned to a small island, and you can only interact with people on that island. Randomly, islands link up and you can migrate/explore the other island. I wanted to play with ideas of cultural exchange on a really small scale. I don't have any regular users yet, but if any of that is interesting to you, you should sign up if only to say hi.
Tech I'm using: Sprites, Cloudflare Workers, SQLite, Litestream, React SSR
Islands only link up every once in a while, so once you land on an island that's it at least for now. When islands link up you'll see a big '?' appear in the page header that you can click on to explore the linked island. There are two islands right now because the populations are low. There's a mechanism for discovering new islands as the island populations grow.
My hope is that by smooshing people together randomly and making it harder to move, people will have a new way (new old way?) of interacting online that has more of the good aspects of offline socialising. But you're 100% right that without a reason to engage it's going to be very quiet. I'm looking at different ways to tackle this that range from "it's a video game now" to something much more subtle.
Not sure if anyone from the deno team is monitoring this forum, but I was trying to stand up a dev-base snapshot and pretty quickly ran into a wall. Is it not currently possible to create a bootable volume from the CLI? https://docs.deno.com/sandbox/volumes/#creating-a-snapshot has an example for the js API, but the CLI equivalent isn't specifying --from and the latest verson of the deno CLI installed fresh from deno.land has no --from option. Is the CLI behind, here? Or is the argument provided some other way?
It's working now, thanks. While I've got your attention, it was a little bit of effort to wrap my head around the APIs when `sandbox create` uses --root AND/or --volume, `snapshot create` uses positional args <volumeIdOrSlug> <snapshotSlug>, and `volumes create` uses --from
I know that each of these things is subtly different, but they're similar enough that the bootable snapshot creation workflow (which I expect is a common one) has some sharp edges, since you have to interact with all three APIs at the same time.
Also, the CLI doesn't give a useful error when you try to create a snapshot from a currently attached volume.
Finally, updating a snapshot is more steps than I'd ideally like. I would much rather be able to make changes in a sandbox with a snapshot root and have them persist as a new snapshot. I kind of get why this isn't currently the case, but The volume/snapshot dance feels (for my usecase) like it's missing some abstraction.
That said, now that I've got a snapshot set up it's a nice experience. I've got an alias for `deno sandbox create --root dev --ssh` and I can `claude` in yolo mode without much fear.
See also Sprites (https://news.ycombinator.com/item?id=46557825) which I've been using and really enjoying. There are some key architecture differences between the two, but very similar surface area. It'll be interesting to see if ephemeral + snapshots can be as convenient as stateful with cloning/forking (which hasn't actually dropped yet, although the fly team say it's coming).
Will give these a try. These are exciting times, it's never been a better time to build side projects :)
Sprites aren't ephemeral. They're like deli cups: "semi-disposable". You keep them around as long as you feel like, and you don't feel bad about throwing them away.
These really blew me away. I have a theory that he recorded the narration and dialogue of each character individually and then it was all edited together - it seems impossible to switch back and forth between such incredible character deliveries on the fly. Or perhaps this is just how that kind of work is done. Regardless, an amazing job.
How is it possible that beef, dairy, and chicken are front and center while Lentils, Tofu (or even just soy), Chickpeas, Nutritional Yeast, Broccoli, etc are all left off? Why do they arbitrarily split "protein" and "fruit/veg" given that most/all of the most protein dense foods are vegetables/legumes? Steak is a terrible source of protein (in terms of nutrient density). Immediately pretty suspicious.
There are nuts and legumes there in the bottom left.
So funny to see people reflexively defend those things being left off because it confirms their own beliefs. A deeper inspection of the actual guidelines has them being very fair to plant proteins:
> Consume a variety of protein foods from animal sources, including eggs, poultry, seafood, and red meat, as well as a variety of plant-sourced protein foods, including beans, peas, lentils, legumes, nuts, seeds, and soy.
The thing is... the pyramid is just a graphic, the actual words give more context.
It's not just a personal belief that plant sources are, on the whole, better from a health perspective.
Since we're talking about the actual wording of the report, it admitted the significance of previous reports deciding to order plant foods before animal products. That is reversed in this most recent report, and very intentionally, which they make clear. They also pretend that the health effects of saturated fat intake are still fuzzy, as if the evidence doesn't heavily point towards it being detrimental.
If anyone is holding to unshakeable beliefs and unwilling to consider evidence, it's the shoddy scientists (many with meat-industry related conflicting interests) that wrote the report.
Kinda of wild that Dairy got its own section in that document as a proscribed thing to eat.
There are plenty of lactose intolerant people. These people can meet their nutritional needs without dairy. (For Calcium: via Sardines, leafy greens, Tofu, etc.)
I guess what I'm lamenting is the missed opportunity to highlight that many vegetables e.g. broccoli are an excellent protein source as well as other important nutrients. It gives you additional flexibility when meal planning. There's a common misconception (at least in my circles) that protein => animal protein which isn't always useful for planning a balanced meal.
Broccoli has 2.8g of protein per 100g. Beef has 26g per 100g, and chicken has 27g. If you're trying to get protein, broccoli isn't going to do much, and I think it's good that the government is being honest about that. A chart that listed broccoli as a major source of protein would be misleading. Broccoli is a good source of many nutrients, and the chart calls it out as such, but it is not an effective source of protein.
If you compare protein per kJ instead, broccoli has 0.021g protein per kJ whereas lean beef mince has 0.028g per kJ. Much more similar. Although of course you would need food that is higher density protein as well so you don't have too much volume to eat.
But that is a kind of silly way to compare. Broccoli isn't very filling _and_ it doesn't have very much protein in it. That doesn't change the fact that it lack protein.
The question is if I'm preparing a meal that I want to be filling, healthy, and energizing, how should I do it. Broccoli isn't a good answer to the protein part of that question.
Normalising by mass is a poor way to assess food's protein content since different foods have greatly different water contents. E.g. beef jerky has much higher protein per 100g than beef largely because it's dried (admittedly, probably also because they use leaner cuts)
> I guess what I'm lamenting is the missed opportunity to highlight that many vegetables e.g. broccoli are an excellent protein source as well as other important nutrients
I can see why you would expect something like that from this administration, but surprisingly the linked webpage seems to be based in fact.
Broccoli are not an excellent protein source from a dietary perspective.
It’s harder to get the target 1-1.6g protein per kg from vegetables, unless you’re consuming beans/pulses which are also high in carbohydrates. Broccoli is not a great protein source, an entire head will give you 10g at most – the average adult would have to eat a dozen+ per day.
You have to consume a very large amount of lentils to make up a healthy amount of protein per day. It’s something like 6 cans of chickpeas vs two chicken breasts per day. I believe you also don’t get a complete amino acids panel like you would with meat which is complete on its own.
> I believe you also don’t get a complete amino acids panel like you would with meat which is complete on its own.
You can challenge beliefs and do a modicum of research, which would easily disprove this false and frankly ridiculous notion, which defies even a rudimentary understanding of plant biology.
I mean I have. There are almost no vegetables that are considered amino acid complete though there are (well known) combinations like legumes (beans/lentils) + rice. But this goes back to my original point of needing a lot of beans to get your protein requirement for the day. In places like India where there are a lot of vegetarians, diary products are heavily used to make up the deficit.
you need to educate yourself better about "basic facts about biology"
they're called essential because humans cannot produce them internally, so we have to consume them (though you could in principle make the same assessment for other animal species, but that's less relevant, unless you're, I don't know, raising cows?)
plants don't eat, but produce organic molecules from raw ingredients (or almost raw, in case of nitrogen), and can produce all amino acids - but in different quantities, so maybe the (parts of) plants you eat don't have all the necessary amino acids.
Now they do produce all the essential amino acids, but in insufficient amounts? Weird how the narrative keeps changing in this thread. A serious lack of scientific knowledge is apparent from people who insist on eating animals. And as always, it is devoid of any backing evidence or credibility other than "trust me, bro, I lift".
From your tone and the fact that you're quoting things nobody in this thread has said, I'm not sure that you are actually interested in hearing any scientific argument. You certainly aren't trying to make one. But I'll try:
The quality of a protein is measured using PDCAAS (Protein digestibility corrected amino acid score). It's a score between 0 and 1 that measures the quality of a protein as a function of digestibility and how well it meets the human amino acid requirements.
It is indeed correct that both lentils and chickpeas (which the original comment you replied to was talking about) have a much lower PDCAAS value of around 0.70. Data on beef varies, but it is generally considered to be a complete protein with a PDCAAS score above 0.90.
Instead of accusing "people who insist on eating animals" of lacking scientific knowledge, it would have been much more helpful to point out that the highest quality proteins on the PDCAAS scale are almost universally vegetarian or vegan: eggs, milk, soy, and mycoprotein all have higher scores than beef, chicken, or pork.
I believe the person you’re responding to is a vegan (from other comments) so the “amino complete” alternative of eggs and dairy you’re suggesting don’t fit the bill of requirements for his arguments either which leaves soy. Mycoprotein has plenty of controversy around it regarding heavy metals and health issues from the fact that it’s highly processed. Soy has a lot of phytoestrogens so it’s not a great candidate to consume large amounts of.
> Soy has a lot of phytoestrogens so it’s not a great candidate to consume large amounts of.
The buffoonery continues. These irrational statements are straight out of the meat industry playbook - of course again lacking in any credible citations. And all you had to do was spend even 5 seconds reading a public encyclopedia to avoid this embarassment https://en.wikipedia.org/wiki/Phytoestrogen#Effects_on_human...
You seem to have a somewhat decent grasp of the facts, but honestly, if you don’t work on your tone, your posts will keep getting downvoted. If you like to yell at scream and call people incompetent, go off to Twitter or some other place that will have you. HN tries to maintain something called tone.
You seem confused. The original claim that plants lack certain amino acids - or that eating them will somehow lead to a protein deficiency - was and is now again thoroughly debunked. The only reason people cling to the notion is to justify their inappropriate diet of animals.
Most protein rich vegetables are legumes and beyond this are also rich in complex carbs. Legumes are in the top 10 food allergies. Not to mention the amino profile of vegetable sources isn't very good.
Beef has ~3x more protein per gram than legumes. It is much more protein-dense than vegetables or legumes.
Similarly, it's a "complete" protein, whereas most vegetables and legumes are missing necessary amino acids.
The downside of beef isn't the "density" of nutrients: the downside is high saturated fat. Chicken breast, though, is similarly high in protein without the saturated fat downside.
> most vegetables and legumes are missing necessary amino acids
In practice, there's no evidence of amino acid deficiency in vegans/vegetarians except ones that restrict even further (potato diet, fruitarians, etc)
https://pmc.ncbi.nlm.nih.gov/articles/PMC6893534/
Besides the ever-popular soybean being a complete protein, if you have normal variety in your diet, it's just not something you have to worry about.
>In practice, there's no evidence of amino acid deficiency in vegans/vegetarians
That is not what your linked article says. It says there is not evidence of protein deficiency, and the deficiency of amino acids is overstated. Not that there is no deficiency.
And vegan/vegetarian health is really a 2nd order variable here. Vegans and vegetarians could have massive amino acid surpluses and it remains a fact that vegetable proteins lack useful amino acids that meat has. Maybe the vegetarians are eating lots of eggs. Maybe they are taking lots of supplements. Maybe they are actually eating meat despite calling themselves vegans and vegetarians. It doesn't matter. There really is no disputing the fact about the composition of meat/vegetable protein.
> a fact that vegetable proteins lack useful amino acids that meat has.
This isn't a problem since you only need nine essential amino acids and they are present in adequate quantity in various vegetables and shrooms. The others are synthesized by ones body.
Only if those vegetables or shrooms were grown in natural sunlight (no greenhouse plastic/glass involved) and in a soil with abundant minerals, macronutrients, and high brix value.
The fat is an excellent source of energy though and it's very hard to get fat by eating fat because it's essentially hormonally inert. I.e. eating fat doesn't precipitate insulin which is the hormone that enables body fat accumulation.
So the problem with steak isn't the steak itself it's the "steak dinner" where the meat comes with sides such as french fries and drinks such as beer.
> The downside of beef isn't the "density" of nutrients: the downside is high saturated fat.
There are other downsides to beef .. such as the batshit crazy use of ecosystems and resources required to produce it at industrial scale.
Got a (beef) cow roaming in your yard, somehow getting by on whatever grows out of the ground? Enjoy your steak! Generating 6x the calories via a water-intensive cover crop to feed the cow so you can eat it later? Just say no.
This is orthogonal to nutritious eating habits; I don't think the food pyramid should lie about nutrition due to ecological concerns. (I do think the food pyramid should be a little more concerned about saturated fat than it is, though — which is why I called out chicken as an alternative, and elsewhere also mentioned fish.)
Worth noting that like amino acids there are essential fatty acids as well, and most people have poor nutrition there... red meat isn't "only" saturated fat, but a fairly balanced fatty acid profile. You can have too much, but in moderate cuts it isn't too bad.
I usually suggest around 0.5g fat to 1g protein as a minimal, higher if keto/carnivore.
That's true, although fish has a better balance of essential fatty acids than red meat. Although, oddly enough, wagyu has a (much) better fatty acids profile than other types of beef, so you can justify the occasional wallet splurge on health grounds!
Steak is actually an excellent source of protein (and fat, if you get the fattier steak as you should).
Just because vegetables, lentils or nuts contain protein it doesn't mean it's the same/equivalent to the protein in an animal product.
Meat is actually super easy for humans to digest and it has no downsides to it. All vegetables on the other side contain plenty of anti-nutrients such as folate and oxalates.
Everything in human body, skin, connective tissues, tendons, hair, nails, muscles is essentially built out of protein and collagen. Fats are essential for hormone function.
> Meat is actually super easy for humans to digest and it has no downsides to it.
In moderate amounts, sure. But frequently eating red meat (more than two or three servings a week) is terrible for you. There's "a clear link between high intake of red and processed meats and a higher risk for heart disease, cancer, diabetes, and premature death": https://www.health.harvard.edu/staying-healthy/whats-the-bee...
Not to mention how high heat cooking of meat, which is common for a steak via frying, brings health risks from Advanced Glycation End products (AGEs).
AGEs are also present in vegetables and legumes, but certain meats like bacon contain unbelievable amounts relative to other foods. (Interestingly: Rice contains almost no AGE's.)
"We put a bunch of meat derived products with high amounts of artificial additives together with actual meat and then concluded that meat is the problem"
I really take issue with studies like this that put meat and meat products together.
Unprocessed meat is what humans what have been eating for hundreds of thousand of years.
Meat products are commercial new inventions and contain stuff like preservatives, volume expanders, flavor enhancers and coloring agents. They also typically contain added sugars, sodium, malto dextrin, corn syrup.
One can't seriously put these together and call them the same, make a study where participants might be eating SPAM and then conclude that "red meat is bad".
Given the choice between "Domino's vegetarian pizza", "IKEAs meatballs" and "steak that is fried,salted and peppered" which one do you think will be the healthiest option?
It's a good point, and maybe Broccoli isn't then as compelling as something like tofu, which contains nearly as much (and nearly as bio available) protein/calorie as lean steak.
I guess I'd challenge the 'no downsides' claim. Few people stick to super lean grass-fed cuts, and the picture on the site is even a ribeye steak :P
The protein density (g/kcal) of a ribeye steak is basically the same as tofu (I think like 14g/100kcal vs 11g/100kcal in tofu)
I know I'm moving the goal posts slightly (I admit I didn't know about bio availability, and see now that I have more to read up on e.g. Broccoli), but am learning as I discuss rather than arguing a fixed point.
Bioavailabilty is a bit of a non-issue. It's measured as if the food you are measuring is the only food you eat. So if it is slightly low on one amino acid, the "bioavailabilty" drops, but noone eats like that. Once combined with other foods, the total "bioavailabilty" tends to increase.
The bigger problem is nutritional density. I tried meeting the 1-1.5 g/kg protein level through a vegetarian whole grains diet and it's a lot of flipping food. Equivalent of like 3kg of chickpeas a day to make it.
It was definitely eye opening on the sort of ancient benefit of meat. It's really hard to reach your muscular potential without it.
An adult who weighs 75 kg, so is targeting about 75 grams of protein intake per day, would only need to eat 833 grams of cooked chickpeas (which are 9% protein by weight) to get there. That is indeed a lot of chickpeas! But a lot less than you claimed, and you probably shouldn't be getting all your protein from chickpeas anyway.
You're probably talking about dry weight. My can says 6g protein / 130 g. I'm about 100kg and to hit the 1.6 g protein/kg I need 160g of protein. 6g/130g * 3500 g is 161 g of protein.
- Canned, drained and rinsed: 7g protein / 100g [1]
- Boiled: 9g protein / 100g [2]
Not sure what explains the discrepancy (though the second number is much older), but both are considerably higher than what your can says. Sure you aren't reading a per-serving amount?
> How is it possible that beef, dairy, and chicken are front and center while Lentils, Tofu (or even just soy), Chickpeas, Nutritional Yeast, Broccoli, etc are all left off?
To quote famed businessman and philosopher Eugene Krabs: "Money."
Not the poster, but, usually what people are referring to is all the other stuff that comes along.
Per calorie beef and broccoli are actually surprisingly similar, but broccoli comes with fiber, calcium and vitamin C, while beef comes with saturated fat.
Of course, broccoli is not very calorie dense, so you would need to eat a lot.
More realistically, tofu, which has about as much protein per calorie (and almost as much per gram) as middling lean beef. But has half the saturated fat, more iron, more calcium, and fibre.
You just get more good stuff, and less bad stuff with veg.
Tofu being displeasurable is funny to me because it literally has no taste and texture by default. It becomes whatever you put it in or how you cook it. You want crunchy? You got it. Puree? Sure. Sweet? Fine. Salty? Spicy? Tangy? Easy.
People just don't want to actually put in the effort to prepare it.
My problem is that I just can't get it to take up any of the flavour. I can marinate it for days, and the marinade will still just be a superficial layer on top of a piece of tofu which, itself, always remains completely unfazed and tasteless.
It's not a problem for saucy dishes like a curry, but even experimenting with friends and borderline "molecular cuisine" techniques I have never once managed to flavour tofu itself :(
yeah that can be very difficult. I think you should aim to _season_ the tofu (i.e. salt -- or slight umami with soy sauce), but your primary flavour should still come from a sauce that's on it. I really like sticky sauces that cling to tofu like buffalo sauce or sugary, sticky glazes.
Beef and chicken does not cause cancer anymore than anything else does. It is an insane take that regular food causes cancer in any level that should be worrisome. Don't cite the studies where they grouped frozen pizza in the same category as beef.
As a flexitarian, I've had to think quite a lot about how to get enough bioavailable protein while moderating my carb consumption and digestive upset due to beans, and to do so in a sustainable manner factoring in convenience and lack of leisure. I certainly won't recommend anything but lean meat and dairy as protein staples to people who aren't used to watching what they eat.
> How is it possible that beef, dairy, and chicken are front and center while Lentils, Tofu (or even just soy), Chickpeas, Nutritional Yeast, Broccoli, etc are all left off?
Possibly because those foods are culturally un-American or something silly like that
> most/all of the most protein dense foods are vegetables/legumes
Are you abusing "dense" to mean calories over calories, rather than the expected calories over weight measure? Even a cursory search shows the latter to be untrue. The former is disingenuous, because despite "density", people do not eat kilograms of broccoli daily to hit minimum-viable protein targets.
I do regret mentioning Broccoli because it seems to have become a bit of a distraction from my original point, which was that getting enough protein from a varied diet actually isn't that hard once you start to notice how much protein is in certain common veg. I'm not totally sure I understand where the mentality that all your protein has to come from a single source in isolation comes from, but suspect representations like this pyramid are at least partly to blame.
Agree that g/kcal isn't perfect but g/g has its own corner cases like water content skewing things badly (e.g. dried spirulina is 57% protein by weight but you'd never eat more than like a gram in a serving). I never meant to suggest that people should be eating broccoli _in place of_ turkey, only that by _de-emphasising_ the protein content of many vegetables in favor of animal proteins, the graphic encourages meal planning that must always contain an animal protein. More insidiously, in my experience at least, it blurs the line between the nutrition content of different animal proteins ("I have my veg I just need 'a protein' now") which leads to more consumption of red meat regardless of quality.
The graphic that I wish someone would make is the 'periodic table of macro nutrients' that positions foods along multiple dimensions at once but I don't know how you would actually do it in just two dimensions.
> Steak is a terrible source of protein (in terms of nutrient density).
At 23g/100g, lean beef has a very high protein/weight ratio. Similar to chicken and turkey breast and exceeded only by canned tuna and processed protein isolates like soy protein isolate, whey protein isolate, and wheat gluten. For comparison, protein content of firm tofu, lentils, and chickpeas is much lower, at 14g/100g, 9g/100g, and 8.5g/100g, respectively. They all contain a lot more carbs per 100g than lean beef.
Further, lean beef contains a full and balanced amino acid profile, which lentils, tofu, chickpeas, soy protein isolate, and wheat gluten does not. It's an excellent food. However there is evidence that charred red meat and red meat containing nitrites is associated with a slight increase in colorectal cancer, so people should be consuming minimally processed red meat where possible, as per the guidance.
Replying to my own comment because I've had some more time to look through the scientific foundation document. In particular, this was an illuminating section (and maybe hinting at where the 'war on protein' language comes from)
> The DGAs recommend a variety of animal source protein foods (ASPFs) and plant
source protein foods (PSPFs) to provide enough total protein to satisfy the minimum
requirements set at the Recommended Dietary Allowance (RDA) of 0.8 g/kg body
weight for adults and to ensure the dietary patterns meet most nutrient needs [3, 4].
However, over the past 20 years, an extensive body of research has underscored the
unique and diverse metabolic roles of protein, and now there is compelling evidence
that consuming additional foods that provide protein at quantities above the RDA may
be a key dietary strategy to combat obesity in the U.S (while staying within calorie limits
by reducing nutrient-poor carbohydrate foods).
Instead of incorporating this approach, the past iterations of the DGAs have eroded
daily protein quantity by shifting protein recommendations to PSPFs, including beans,
peas, and lentils, while reducing and/or de-emphasizing intakes of ASPFs, including
meats, poultry, and eggs. The shift towards PSPFs was intended to reduce adiposity
and risks of chronic diseases but was primarily informed by epidemiological evidence on
The Scientific Foundation for the Dietary Guidelines for Americans, 2025–2030: Appendices | 350
dietary patterns, even in some cases when experimental evidence from randomized
controlled trials (RCTs) was available to more specifically inform this recommendation.
Another key aspect that DGA committees have inadequately considered are the nutrient
consequences when shifting from ASPFs to PSPFs. ASPFs not only provide EAAs, they
also provide a substantial amount of highly bioavailable essential micronutrients that are
under-consumed. Encouraging Americans to move away from these foods may further
compromise the nutrient inadequacies already impacting many in the U.S., especially
our young people.
Compounding this is the recent evidence highlighting the fallacies of using the
unsubstantiated concept of protein ounce equivalents within food pattern (substitution)
modeling, leading to recommended reductions in daily protein intakes and protein
quality since ASPFs and PSPFs are not equivalent in terms of total protein or EAA
density. Given that 1) there is no Tolerable Upper Intake Level (UL) for dietary protein
established by the Dietary Reference Intakes (DRIs) and 2) consuming high quality
ASPFs above current recommendations has shown no negative health risks in high
quality RCTs, it’s unclear as to why previous DGAs encouraged shifts in protein intake
towards limiting high quality, nutrient dense ASPFs. It's essential to evaluate the
evidence to establish a healthy range of protein intake and to substantiate whether or
not limiting ASPFs is warranted and/or has unintended consequences.
An alternative approach that may be more strongly supported by the totality of evidence
is the replacement of refined grains with PSPFs like beans, peas, and lentils. Given
their nutrient dense profile (e.g., excellent source of fiber, complex carbohydrates, &
folate, etc.; good source of protein) nutrient dense PSPFs complement but do not
replace the nutrients provided in ASPFs (i.e., excellent source of protein, vit B12, zinc,
good source of heme iron, etc.). By including high quality, nutrient dense ASPFs as the
primary source of protein, followed by nutrient dense PSPFs as a replacement for
nutrient-poor refined grains, a higher-protein, lower-carbohydrate dietary pattern can be
achieved which likely improves nutrient adequacy, weight management, and overall
health. -- https://cdn.realfood.gov/Scientific%20Report%20Appendices.pd... Appendix 4.9
No vegetable is as protein dense as actual meat in its natural form.
Ruminant meat is absolutely one of the best bioavailable forms of a mostly complete amino acid profile, though eggs and dairy is more complete with differing ratios depending on form/feed.
As to lentils, tofu, chickpeas etc. They're fine for most people in moderation, but they are also relatively inflammatory and plenty of people have digestive issues and allergies to legumes (I do), soy is one of the top 10 allergens that people face. While almost nobody is allergic to ruminant meat.
This is such a bonkers line of thinking, I'm so intrigued. So a particular model will have an entire 'culture' only available or understandable to itself. Seems kind of lonely. Like some symbols might activate together for reasons that are totally incomprehensible to us, but make perfect sense to the model. I wonder if an approach like the one in https://www.anthropic.com/research/tracing-thoughts-language... could ever give us insight into any 'inside jokes' present in the model.
In some sense it's more human than a model trained with no RL and which has absolutely no exposure to its own output.
We have our own personal 'culture' too-- it's just less obvious because its tied up with our own hidden state. If you go back and read old essays that you wrote you might notice some of it-- that ideas and feelings (maybe smells?) that are absolutely not explicitly in the text immediately come back to you, stuff that no one or maybe only a spouse or very close friend might think.
I think it may be very hard to explore hidden subtext because the signals may be almost arbitrarily weak and context dependent. The bare model may need only a little nudge to get to the right answer and the you have this big wall of "reasoning" where each token could carry very small amounts of subtext that cumulatively add up to a lot and push things in the right direction.
THANK YOU so much for this. I'd spend some time researching how to disable YouTube shorts only to come to the conclusion that I had to use a different client (I picked NewPipe, but faced frequent issues with playback and I ended up having to click out to YouTube 90% of the time anyway)
I've disabled YouTube history now and it's almost exactly what I'd hoped for. Shorts are disabled, no 'feed', no suggestions, just my subscriptions. I owe you a beer.
My pleasure, I went down that rabbit hole myself until I saw a comment in reddit that helped me to make youtube the way I like it. I am just passing it forward and glad to hear it helped you.
I'm not in any way qualified to have a take here, but I have one anyway:
My understanding is that entropy is a way of quantifying how many different ways a thing could 'actually be' and yet still 'appear to be' how it is. So it is largely a result of an observer's limited ability to perceive / interrogate the 'true' nature of the system in question.
So for example you could observe that a single coin flip is heads, and entropy will help you quantify how many different ways that could have come to pass. e.g. is it a fair coin, a weighted coin, a coin with two head faces, etc. All these possibilities increase the entropy of the system. An arrangement _not_ counted towards the system's entropy is the arrangement where the coin has no heads face, only ever comes up tails, etc.
Related, my intuition about the observation that entropy tends to increase is that it's purely a result of more likely things happening more often on average.
Would be delighted if anyone wanted to correct either of these intuitions.
>purely a result of more likely things happening more often on average
according to your wording, no. if you have a perfect six sided die (or perfect two sided coin), none/neither of the outcomes are more likely at any point in time... yet something approximating entropy occurs after many repeated trials. what's expected to happen is the average thing even though it's never the most likely thing to happen.
you want to look at how repeated re-convolution of a function with itself always converges on the same gaussian function, no matter the shape of the starting function is (as long as it's not some pathological case, such as an impulse function... but even then, consider the convolution of the impulse function with the gaussian)
> My understanding is that entropy is a way of quantifying how many different ways a thing could 'actually be' and yet still 'appear to be' how it is. So it is largely a result of an observer's limited ability to perceive / interrogate the 'true' nature of the system in question.
When ice cubes in a glass of water slowly melt, and the temperature of the liquid water decreases, where does the limited ability of an observer come into play?
It seems to me that two things in this scenario are true:
1) The fundamental physical interactions (i.e. particle collisions) are all time-reversible, and no observer of any one such interaction would be able to tell which directly time is flowing.
2) The states of the overall system are not time-reversible.
The temperature of an object is a macroscopic property basically depending on the kinetic energy of the matter within it, which in a typical cup of water varies substantially from one molecule to the next. If before you could guess a little bit about the kinetic energy of a given water molecule based on whether it is part of the ice or not, after melting and sufficient time to equilibrate the location of a particular molecule gives you no additional information for estimating its velocity.
It's tricky when you think of a continuous system because the "differential entropy" is different (and more subtle) than the "entropy". Even if a system is time-reversible, the "measure" of a set of states can change.
For example: Say I'm at some distance from you, between 0 and 1 km (all equiprobable). Now I switch to being 10x as far away. This is time-reversible, but because the volume of the set of states changed, the differential entropy changes. This is the kind of thing that happens in time-reversible continuous systems that can't happen in time-reversible discrete systems.
Isn't that kind of what we want entropy to capture though? If a particle darts off into the distance then in theory it might be time reversible, but in practice it's not so simple. If the particle escapes the gravitational pull, the only way it can come back is if it bumps into some other object and pushes that object away. So things will inevitably spread out more and more creating an arrow of time.
This can then be related to the big bang, and maybe it could be said that we are all living of the negentropy from that event and the subsequent expansion.
Getting different entropy values based on choice of units is a very nasty property though. It kinda hints that there is one canonical correct unit (plank length?)
I have yet to see differential entropy used successfully (beyond its explicitly constructed-for purpose for calculating channel capacity). Similar to your thought experiment is the issue that the differential entropy value depends on your choice of unit system. Fundamentally the issue is that you cant stick a quantity with units into a transcendental function and get meaningful results out
Yeah, it's quite disturbing that the differential entropy (unlike the discrete entropy) depends on the units. Even worse, the differential entropy can be negative!
Interestingly, the differential KL-divergence (differential cross-entropy - differential entropy) doesn't seem to have any of these problems.
Not sure I'm going to get any traction here, but seeing as the 'support' button in the website is greyed out, not sure where else to post this. I got double-charged for my subscription, and I'm hoping you folks can help me get a refund on the extra charge. I was going through the subscription flow and got an error. The website allowed me to subscribe again so I did, assuming the first one hadn't gone through. I now have two charges on my card.
I have a feeling this particular brand of hair splitting is going to be an interesting fixture in the history books.