Soylent costs more than my current diet, limiting gains
it is a priori highly likely to fail since we know for a fact that severe nutrition deficiencies can be due to subtle & misunderstood factors (see: the forgetting of scurvy cures) and that nutrition is one of the least reliable scientific areas
his work is even more likely than that to have problems because he hasn’t consulted the existing work on food replacements (yes, it’s a thing; how exactly do you think people in comas or with broken jaws get fed?)
given #2, the negative effects are likely to be subtle and long-term means that on basic statistical power grounds, you’ll want long and well-powered self-experiments to go from ‘crappy self-experiment’ to ‘good self-experiment’*
given the low odds of success (#2-3), the expensive powerful self-experiments necessary to shift our original expectations substantially due to long-term effects and subtlety (#4), and the small benefits (#1), the VoI is low here
my other self-experiments, in progress and planned, suffer from many fewer of Soylent’s defects, hence have reasonable VoIs (Specifically: I am or will be investigating Noopept, melatonin, magnesium l-threonate & citrate, coluracetam, meditation, Redshift, and lithium orotate.)
VoI current/planned self-experiments (#6) > VoI Soylent cloning/tweaking (#5)
hence, the opportunity cost of Soylent is higher than not, so I will continue my existing plans
* although see my reply to Qiaochu, at this point Rob isn’t even at the ‘crappy’ level
EDIT: as of June 2015, I would amend my list of complaints to de-emphasize #3 as it seems that Soylent Inc has revised the formulation a number of times, run it by some experts, and has now been field-tested to some degree; most of my self-experiments in #6 have since finished (right now the only relevant ones are another magnesium self-experiment, trying to find the right dosage, and nonrandomized bacopa ABA quasiexperiment); and for point #1, between increasing the protein in my diet and official Soylent lowering prices, now Soylent is more like 2x my current food expenditures than 3x+.
it is a priori highly likely to fail since we know for a fact that severe nutrition deficiencies can be due to subtle & misunderstood factors (see: the forgetting of scurvy cures) and that nutrition is one of the least reliable scientific areas
While this is true, I would expect that for many people, the main risk in Soylent would be one of overdosing on something rather than acquiring a deficiency. Soylent at least tries to provide everything that you need, while many people (me included) optimize their normal diet mainly based on criteria like cost and ease of preparation rather than healthiness. And even if we did optimize it for healthiness, your argument is essentially saying that we might very well screw up and acquire deficiencies even if we tried to ensure that we got everything that we needed.… and that argument can be applied to normal diets just as well as Soylent.
So anyone who uses the “there are lots of subtle ways of acquiring nutrition deficiencies and we might not know everything that one needs” argument against Soylent would first need to show why normal diets would avoid that argument any better.
And even if we did optimize it for healthiness, your argument is essentially saying that we might very well screw up and acquire deficiencies even if we tried to ensure that we got everything that we needed.… and that argument can be applied to normal diets just as well as Soylent.
I disagree. A random basket of foods, while likely deficient in something or other, will also likely change what it’s deficient in over time, while Soylent by definition will be consistently deficient. Even if regular food vs Soylent were both equally harmful, the harm from the regular food may be less due to the variability of it. Statistics analogy borrowing from Jaynes: total error in an sampling estimate can be broken down as random error vs systematic error/bias—but random errors gradually cancel out as the sample size increases, while systematic errors remain the same. Soylent is all systematic error.
So anyone who uses the “there are lots of subtle ways of acquiring nutrition deficiencies and we might not know everything that one needs” argument against Soylent would first need to show why normal diets would avoid that argument any better.
I sympathize with this argument, but the obvious counter-argument is that lots of people have eaten normal diets and have been observed not to, for example, die of scurvy. (On the other hand, they have been observed to, for example, get heart disease.)
I sympathize with this argument, but the obvious counter-argument is that lots of people have eaten normal diets and have been observed not to, for example, die of scurvy. (On the other hand, they have been observed to, for example, get heart disease.)
That’s true. But then again, once you consider that “normal diets” is really composed of countless of different combinations of foods ranging from “fast food only” to “making a constant effort to be trying out new foods all the time”, you could also use this as an argument for Soylent being probably safe. As in, “out of all the countless possible combinations of nutritional intakes that people live on, most don’t lead to anybody dying of scurvy, so if we specifically construct one new diet for the express purpose of providing everything that one needs, it doesn’t seem like it should kill you if all those diets that weren’t constructed with that in mind don’t kill you”.
Only if you are able to track the deficiency back to its cause. To reuse scurvy, how many realized that their deficiency was of fresh fruits and vegetables? As opposed to bad air or bacterial poisoning or whatever… If you felt the symptoms of rabbit starvation but had never heard of it or been told about it, would you realize what the problem was in your diet before you happened to eat something fatty and noticed your vague hunger was finally satisfied?
So anyone who uses the “there are lots of subtle ways of acquiring nutrition deficiencies and we might not know everything that one needs” argument against Soylent would first need to show why normal diets would avoid that argument any better.
We are adapted to obtain nutrients from food. Since we currently lack a good understanding of exactly what properties of food are nutritionally relevant, it seems unwise to replace natural food with artificial food.
Yes, processed foods are quite unnatural. But Soylent is even less natural than processed foods, so this is irrelevant in the present context.
From an evolutionary standpoint, legumes, milk, and grains are “artificial” food, at least for humans. Agriculture is a recent thing. Would you also endorse the Paleolithic diet movement?
(I do actually endorse the paleolithic diet as probably optimal at the moment and I agree with your central point—I just want to point out that even unprocessed modern diets are already rather unnatural.)
Which is an order of magnitude less than the 200,000 years that we’ve been anatomically modern—although who is to say that they didn’t gather wild grains back then, too.
Of course, even 10,000 years is more than enough time for evolution to change us.
it is a priori highly likely to fail since we know for a fact that severe nutrition deficiencies can be due to subtle & misunderstood factors (see: the forgetting of scurvy cures) and that nutrition is one of the least reliable scientific areas
I think you’re wrong about that. We have modern chemistry and we have animal experimentation at scale, which means that we can feed animals highly-refined diets to determine whether any essential nutrients are missing from our models. It would be extremely surprising if there were a vital nutrient we didn’t know about.
On the other hand, there are other failure modes besides forgetting a nutrient, like using an inactive or degraded input, contamination, or for that matter, making half the calories sugar. (Which they, um, did.) I really want a correctly-executed version of Soylent, but I won’t be eating anything from the first batches, because these guys really don’t fill me with confidence.
I agree food-replacements should be doable in theory and that the existing products shouldn’t be too terrible, but Soylent does not seem to be drawing well on the existing knowledge.
I really want a correctly-executed version of Soylent, but I won’t be eating anything from the first batches, because these guys really don’t fill me with confidence.
Personally, I’d like a good Soylent too. It’d be useful for my self-experiments, since it’d help tamp down variability from my diet and increase the statistical power. But Rhinehart is doing it all wrong.
I agree with most of what you are saying, however #2 is likely to be mitigated by his not going on a soylent-only diet. Thus there is a fair chance that many subtle overlooked deficiencies in the product will be masked by the “normal” meals he still eats fairly regularly. In your scurvy example, the minimum level of Vit C required (8-10 mg per day) is far lower than what you get from a typical diet (some 10 times that, apparently), so even if he completely removed it from his product, he’d probably get enough of it from his infrequent non-soylent meals. Though his example of forgetting sulfur is a bit worrying and is evidence against this.
Sure. But I’d point out that this observation (that you can hedge your bets) cuts against Soylent as well: while consuming regular food to limit your downside from deficiencies in Soylent should work for overprovisioned substances, you’re also limiting your upside since the more regular food you consume the less Soylent you must be consuming.
Depends on what you mean by “optimistic that this will work.” Presumably Eliezer at least thinks this is positive expected value (and so do I). That doesn’t have to be because he assigns a high probability to it having positive value, it could be because he assigns a moderate probability to it having moderately high positive value or because he assigns a low probability to it having extremely high positive value, etc.
When computing the expected value, keep in mind that Soylent displaces other food, so the actual cost (assuming the project meets its funding goals) is not $65 but $65 minus however much Eliezer would otherwise have spent on food in a week. For me, and I suspect for Eliezer as well, this number is more than $65, so Eliezer can think that the expected value of replacing his food with Soylent is somewhat negative and still think it’s a good idea to try it for a week. Soylent instead of other food also saves food preparation time in addition to saving money.
Depends on what you mean by “optimistic that this will work.”
I mean buying in at the $65 for a week level.
That doesn’t have to be because he assigns a high probability to it having positive value, it could be because he assigns a moderate probability to it having moderately high positive value or because he assigns a low probability to it having extremely high positive value, etc.
I’m trying to understand the “moderate probability” part. EY’s been on so many non-working diets; this is evidence against a dietary solution working, unless there’s a reason why Soylent isn’t in that reference class.
When computing the expected value, keep in mind that Soylent displaces other food, so the actual cost (assuming the project meets its funding goals) is not $65 but $65 minus however much Eliezer would otherwise have spent on food in a week.
Have you not come across “meal replacement diet” (which may be partial or total) until now? There is a bunch of articles about it in Google Scholars, not to mention the popular media.
I don’t understand. Soylent contains maltodextrin, oat powder, whey protein from milk, olive oil, various vitamins and minerals, whereas, Slim-Fast, for example, contains milk, milk protein concentrate, sugar, maltodextrin, canola oil, various vitamins and minerals, etc. How is one more like regular food than the other?
Considering the amount of time, effort, money, and pain you have been or are willing to put in to decrease your fatness, I want to make sure that you’ve actually considered what your evidence is for whatever benefits and costs you perceive in decreasing your fatness. I haven’t looked into studies in detail, but I think even reflecting on the discourse surrounding fat has a large effect on one’s probability estimate for ‘fat is evil’. By a noticeable (though not necessarily decisive) margin, I find the most plausible explanation for what little I know about fat to be that the world is crazy, people are mad and bigoted, fatness in itself does not on average cause any actionable (i.e. calling for large intervention) significant net loss of health, the medical community has failed to convincingly demonstrate such massive ill effects after controlling for other more plausible causes despite trying extremely hard to because it is privileging a false hypothesis, and that this research agenda is both motivated by and feeds into the aforementioned societal craziness.
(I claim that I’m not counter-other-optimizing-Eliezer_Yudkowsky / epistemic other-optimizing, but I suspect it’d be epistemic other-optimizing to insist you believe that.)
Low, since I do my cooking and cleaning in downtime when I can’t bear to read any more or do something productive; and I also favor recipes like giant crockpots of soup which are both dead-easy to make and consume.
Interesting, googling around a bit it looks like it is basically soybean oil, whey, and dextrose with vitamin powders. So pretty much the same as Soylent. I guess worries about bioavailability are overblown given that coma patients survive indefinitely, but then again, their mixture is adjusted daily based on blood work.
He has; for me, cooking is about three minutes of effort spread out over the course of fifteen minutes each day.
The main benefit to a Soylent-style diet is that you get all the micronutrients and so on that are normally locked away inside vegetables, which take an inordinate amount of time to prepare and consume, in one quick drinkable source, without any concerns about pesticides or fungi or so on.
that are normally locked away inside vegetables, which take an inordinate amount of time to prepare and consume
Use frozen pre-sliced vegetables. They’re usually competitive with, if not cheaper than, fresh produce; you save a lot of time preparing them; they taste just as good to me (and better, in instances where it takes a long time to eat your way through a purchase and the fresh vegetables are, shall we say, less than fresh by the time you eat them). You can easily dump them into crockpot recipes, or you can just put them in a bowl, microwave with some herbs/spices and butter, and reheat as necessary.
Agreed. I use frozen vegetables for anything I cook that has vegetables in it, but that’s pretty rare. Mostly I eat raw sliced sweet potatoes (where the time is in the peeling and slicing) and microwaved kale (which has gotten much less time consuming now that I buy the precut and prewashed version, rather than bunches).
I dislike kale, so no comments there, but why would you peel sweet potatoes? The skin tastes fine and I assume like regular potatoes has a lot of the nutrients in it. When I harvest my sweet potatoes, I just slice them.
Factors why I have not and probably will not:
Soylent costs more than my current diet, limiting gains
it is a priori highly likely to fail since we know for a fact that severe nutrition deficiencies can be due to subtle & misunderstood factors (see: the forgetting of scurvy cures) and that nutrition is one of the least reliable scientific areas
his work is even more likely than that to have problems because he hasn’t consulted the existing work on food replacements (yes, it’s a thing; how exactly do you think people in comas or with broken jaws get fed?)
given #2, the negative effects are likely to be subtle and long-term means that on basic statistical power grounds, you’ll want long and well-powered self-experiments to go from ‘crappy self-experiment’ to ‘good self-experiment’*
given the low odds of success (#2-3), the expensive powerful self-experiments necessary to shift our original expectations substantially due to long-term effects and subtlety (#4), and the small benefits (#1), the VoI is low here
my other self-experiments, in progress and planned, suffer from many fewer of Soylent’s defects, hence have reasonable VoIs (Specifically: I am or will be investigating Noopept, melatonin, magnesium l-threonate & citrate, coluracetam, meditation, Redshift, and lithium orotate.)
VoI current/planned self-experiments (#6) > VoI Soylent cloning/tweaking (#5)
hence, the opportunity cost of Soylent is higher than not, so I will continue my existing plans
* although see my reply to Qiaochu, at this point Rob isn’t even at the ‘crappy’ level
EDIT: as of June 2015, I would amend my list of complaints to de-emphasize #3 as it seems that Soylent Inc has revised the formulation a number of times, run it by some experts, and has now been field-tested to some degree; most of my self-experiments in #6 have since finished (right now the only relevant ones are another magnesium self-experiment, trying to find the right dosage, and nonrandomized bacopa ABA quasiexperiment); and for point #1, between increasing the protein in my diet and official Soylent lowering prices, now Soylent is more like 2x my current food expenditures than 3x+.
While this is true, I would expect that for many people, the main risk in Soylent would be one of overdosing on something rather than acquiring a deficiency. Soylent at least tries to provide everything that you need, while many people (me included) optimize their normal diet mainly based on criteria like cost and ease of preparation rather than healthiness. And even if we did optimize it for healthiness, your argument is essentially saying that we might very well screw up and acquire deficiencies even if we tried to ensure that we got everything that we needed.… and that argument can be applied to normal diets just as well as Soylent.
So anyone who uses the “there are lots of subtle ways of acquiring nutrition deficiencies and we might not know everything that one needs” argument against Soylent would first need to show why normal diets would avoid that argument any better.
I disagree. A random basket of foods, while likely deficient in something or other, will also likely change what it’s deficient in over time, while Soylent by definition will be consistently deficient. Even if regular food vs Soylent were both equally harmful, the harm from the regular food may be less due to the variability of it. Statistics analogy borrowing from Jaynes: total error in an sampling estimate can be broken down as random error vs systematic error/bias—but random errors gradually cancel out as the sample size increases, while systematic errors remain the same. Soylent is all systematic error.
Hmm. I’ll grant that.
I sympathize with this argument, but the obvious counter-argument is that lots of people have eaten normal diets and have been observed not to, for example, die of scurvy. (On the other hand, they have been observed to, for example, get heart disease.)
That’s true. But then again, once you consider that “normal diets” is really composed of countless of different combinations of foods ranging from “fast food only” to “making a constant effort to be trying out new foods all the time”, you could also use this as an argument for Soylent being probably safe. As in, “out of all the countless possible combinations of nutritional intakes that people live on, most don’t lead to anybody dying of scurvy, so if we specifically construct one new diet for the express purpose of providing everything that one needs, it doesn’t seem like it should kill you if all those diets that weren’t constructed with that in mind don’t kill you”.
Not to mention, worst case scenario, if you experience a deficiency, you are still in civilization and may switch back to a normal diet.
Only if you are able to track the deficiency back to its cause. To reuse scurvy, how many realized that their deficiency was of fresh fruits and vegetables? As opposed to bad air or bacterial poisoning or whatever… If you felt the symptoms of rabbit starvation but had never heard of it or been told about it, would you realize what the problem was in your diet before you happened to eat something fatty and noticed your vague hunger was finally satisfied?
The wisdom of nature
Not sure how that applies here, even if we disregard the processed foods that many people live on also being quite unnatural.
We are adapted to obtain nutrients from food. Since we currently lack a good understanding of exactly what properties of food are nutritionally relevant, it seems unwise to replace natural food with artificial food.
Yes, processed foods are quite unnatural. But Soylent is even less natural than processed foods, so this is irrelevant in the present context.
From an evolutionary standpoint, legumes, milk, and grains are “artificial” food, at least for humans. Agriculture is a recent thing. Would you also endorse the Paleolithic diet movement?
(I do actually endorse the paleolithic diet as probably optimal at the moment and I agree with your central point—I just want to point out that even unprocessed modern diets are already rather unnatural.)
Although agriculture is only about 10,000 years old, humans have been gathering and eating wild grains for 30,000.
Which is an order of magnitude less than the 200,000 years that we’ve been anatomically modern—although who is to say that they didn’t gather wild grains back then, too.
Of course, even 10,000 years is more than enough time for evolution to change us.
I think you’re wrong about that. We have modern chemistry and we have animal experimentation at scale, which means that we can feed animals highly-refined diets to determine whether any essential nutrients are missing from our models. It would be extremely surprising if there were a vital nutrient we didn’t know about.
On the other hand, there are other failure modes besides forgetting a nutrient, like using an inactive or degraded input, contamination, or for that matter, making half the calories sugar. (Which they, um, did.) I really want a correctly-executed version of Soylent, but I won’t be eating anything from the first batches, because these guys really don’t fill me with confidence.
I agree food-replacements should be doable in theory and that the existing products shouldn’t be too terrible, but Soylent does not seem to be drawing well on the existing knowledge.
Personally, I’d like a good Soylent too. It’d be useful for my self-experiments, since it’d help tamp down variability from my diet and increase the statistical power. But Rhinehart is doing it all wrong.
I agree with most of what you are saying, however #2 is likely to be mitigated by his not going on a soylent-only diet. Thus there is a fair chance that many subtle overlooked deficiencies in the product will be masked by the “normal” meals he still eats fairly regularly. In your scurvy example, the minimum level of Vit C required (8-10 mg per day) is far lower than what you get from a typical diet (some 10 times that, apparently), so even if he completely removed it from his product, he’d probably get enough of it from his infrequent non-soylent meals. Though his example of forgetting sulfur is a bit worrying and is evidence against this.
Sure. But I’d point out that this observation (that you can hedge your bets) cuts against Soylent as well: while consuming regular food to limit your downside from deficiencies in Soylent should work for overprovisioned substances, you’re also limiting your upside since the more regular food you consume the less Soylent you must be consuming.
Guess we’re all stuck with Soylent then! In for $65.
See also: http://news.ycombinator.com/item?id=5746844
I look forward to your self-experiments...
I’m curious why you’re apparently optimistic that this will work for you, when nothing else has.
Depends on what you mean by “optimistic that this will work.” Presumably Eliezer at least thinks this is positive expected value (and so do I). That doesn’t have to be because he assigns a high probability to it having positive value, it could be because he assigns a moderate probability to it having moderately high positive value or because he assigns a low probability to it having extremely high positive value, etc.
When computing the expected value, keep in mind that Soylent displaces other food, so the actual cost (assuming the project meets its funding goals) is not $65 but $65 minus however much Eliezer would otherwise have spent on food in a week. For me, and I suspect for Eliezer as well, this number is more than $65, so Eliezer can think that the expected value of replacing his food with Soylent is somewhat negative and still think it’s a good idea to try it for a week. Soylent instead of other food also saves food preparation time in addition to saving money.
I mean buying in at the $65 for a week level.
I’m trying to understand the “moderate probability” part. EY’s been on so many non-working diets; this is evidence against a dietary solution working, unless there’s a reason why Soylent isn’t in that reference class.
Obviously.
All of the other diets involve food?
Yep.
Have you not come across “meal replacement diet” (which may be partial or total) until now? There is a bunch of articles about it in Google Scholars, not to mention the popular media.
The meal replacements I’ve seen look a lot more like regular food than Soylent does.
I don’t understand. Soylent contains maltodextrin, oat powder, whey protein from milk, olive oil, various vitamins and minerals, whereas, Slim-Fast, for example, contains milk, milk protein concentrate, sugar, maltodextrin, canola oil, various vitamins and minerals, etc. How is one more like regular food than the other?
Considering the amount of time, effort, money, and pain you have been or are willing to put in to decrease your fatness, I want to make sure that you’ve actually considered what your evidence is for whatever benefits and costs you perceive in decreasing your fatness. I haven’t looked into studies in detail, but I think even reflecting on the discourse surrounding fat has a large effect on one’s probability estimate for ‘fat is evil’. By a noticeable (though not necessarily decisive) margin, I find the most plausible explanation for what little I know about fat to be that the world is crazy, people are mad and bigoted, fatness in itself does not on average cause any actionable (i.e. calling for large intervention) significant net loss of health, the medical community has failed to convincingly demonstrate such massive ill effects after controlling for other more plausible causes despite trying extremely hard to because it is privileging a false hypothesis, and that this research agenda is both motivated by and feeds into the aforementioned societal craziness.
(I claim that I’m not counter-other-optimizing-Eliezer_Yudkowsky / epistemic other-optimizing, but I suspect it’d be epistemic other-optimizing to insist you believe that.)
This did control for lots of stuff.
1 is based on assigning what value to your time?
Low, since I do my cooking and cleaning in downtime when I can’t bear to read any more or do something productive; and I also favor recipes like giant crockpots of soup which are both dead-easy to make and consume.
Do you have any pointers for finding info on #3? Searching around for various related terminology isn’t coming up with much.
Possible starting point: Parenteral nutrition
Interesting, googling around a bit it looks like it is basically soybean oil, whey, and dextrose with vitamin powders. So pretty much the same as Soylent. I guess worries about bioavailability are overblown given that coma patients survive indefinitely, but then again, their mixture is adjusted daily based on blood work.
It’s also given via IV, which means bioavailability via digestion wouldn’t apply.
o.O
My current diet runs me around $2-4 a day; Soylent is ~$10 a day, which is a significant increase. I expect gwern’s numbers are similar.
(I was about to ask about the value of the time spent cooking, but it turns out he’s already answered that.)
He has; for me, cooking is about three minutes of effort spread out over the course of fifteen minutes each day.
The main benefit to a Soylent-style diet is that you get all the micronutrients and so on that are normally locked away inside vegetables, which take an inordinate amount of time to prepare and consume, in one quick drinkable source, without any concerns about pesticides or fungi or so on.
Use frozen pre-sliced vegetables. They’re usually competitive with, if not cheaper than, fresh produce; you save a lot of time preparing them; they taste just as good to me (and better, in instances where it takes a long time to eat your way through a purchase and the fresh vegetables are, shall we say, less than fresh by the time you eat them). You can easily dump them into crockpot recipes, or you can just put them in a bowl, microwave with some herbs/spices and butter, and reheat as necessary.
Agreed. I use frozen vegetables for anything I cook that has vegetables in it, but that’s pretty rare. Mostly I eat raw sliced sweet potatoes (where the time is in the peeling and slicing) and microwaved kale (which has gotten much less time consuming now that I buy the precut and prewashed version, rather than bunches).
I dislike kale, so no comments there, but why would you peel sweet potatoes? The skin tastes fine and I assume like regular potatoes has a lot of the nutrients in it. When I harvest my sweet potatoes, I just slice them.