Hacker News new | past | comments | ask | show | jobs | submit login
Let an algorithm tell you how to eat (nytimes.com)
64 points by petethomas 17 days ago | hide | past | web | favorite | 44 comments

If you're interested in weight loss, I can vouch for this algorithm which has been around a fair while: https://www.fourmilab.ch/hackdiet/online/hdo.html

(I used a way more informal version but the spirit of the exercise was the same.)

This weighted average graph did more to keep me on track than anything else. Sometimes you just need the encouragement to persist through the ups and downs.

Good old Hacker's Diet. Calorie counting with a target adjusted by a moving weighted average of daily weight measurements.

You've had success with it? What did you do to make it informal?

I used his Palm Pilot app which tells you how long ago I first tried it.

I made it informal basically by being lazy. :P If I was more than 2-3kg over my target weight then I'd just eat the minimum calories I can get away with without getting cranky (~1500ish). Once I reached my target weight I totted up crude calorie counts in my head each day (sandwich = 400, biscuit = 100, plate of nachos = 800, that sort of level of detail), keeping the total under my current target. Then I weighed myself every morning, and if I was over my target weight for the whole week, I'd reduce my target calories by 100 or so. If I was under for the whole week, I'd increase my target calories.

It's kept at my target weight (barring a couple of brief lapses of willpower... I'm a bit tubby at the moment!) for most of the last decade. :)

sigh Knowing how to eat healthily isn't a hard problem. The problem is sticking to the diet. But that one's hard, so instead we're constantly solving the first, solved, problem.

I find the easiest way to force myself to eat healthy is to just not have any “undesirable” foods around, planning my grocery trips in great detail before hand, and to cut back on how much I eat out. Cutting off any sources of unwanted foods and focusing on only desirable foods cuts away any of the self control or discipline problems.

People think that will power is not eating the ice cream in the fridge. Very few people can do that. Will power is not buying the ice cream in the first place. It turns out, that is infinitely easier.

The result is you open the freeze a few times, wishing there was ice cream. When there's only frozen chicken, you close the freezer, drink some water, pace around, then get back to whatever you were previously doing.

Wow, thank you for reminding me, I have ice-cream in the fridge. :)

Joke aside, you can only do that if you are alone. I agree in principle, but in a family setting you need to find other changes to routine that result in less food being eaten by you.

Define "eat healthily". Everyone seems to have a different opinion, and whatever you say will be contradicted by someone else.

The most vital overlooked nutrient in the Western diet is controversy. Start from there.

Sorry, I have no idea what you're trying to say.

Indeed, proper diet is no secret, and you don’t even need to do exercise (not that you shouldn’t, but it’s not a strict requirement for weight loss at all). It seems that it’s easier to keep blaming the “solved problem” so yet another product can be sold.

I find the hardest part of eating healthy is at the time of making the choice of what to eat.

A.I could potentially suggest restaurant orders, grocery lists, and recipes. If I had confidence that the system had my best interets "at heart" and it had enough expert knowledge built in I'd probably let it make 95% of my diet choices and count myself lucky to make one less type of decision while improving my health.

I'm working on something just for this, no A.I required. Most of the problems with eating healthy in my opinion is just being prepared with the stuff to make and deciding what to make in advance. Most of the problems I've found with existing "food/recipe sites" is they quickly devolve into "branded advertising", "data collection", and/or "choice overload." Does any site really need over 10,000 choices of "chicken soup" recipes? The other problem with the Internet (there are always more problems with the Internet) is when you search for recipes in the uncurated web, every recipe is prefixed by an intensely personal story about the family you don't even know. Once you find that recipe, it's in a tiny hard to read box (I don't know who made this template) that naturally has been made very hard to copy and paste elsewhere.

Its a terrible user experience.

My idea isn't ready yet, but I'm furiously working on the MVP.

I'd be interested in this.

This is an easy problem to solve, just always eat the same thing. No need to think at the grocery store, no need to decide what your next meal will be. Food is pretty much completely removed from your thought process and incurs no cognitive load.

I'm not just saying this as some rando internet know-it-all, I lost half my body weight calorie counting.

I have to disagree. Everyone knows that they should eat a salad instead of a cake. But what does healthy eating look like normally, when it gets more nuanced, and when you might actually enjoy eating it? Does a healthy meal look like a steak with a fried egg on top and a side of vegetables? Or does it look like fat free yogurt and a Coke Zero?

You'll get a lot of different opinions.

< I had several glucose spikes as high as 160 milligrams per deciliter of blood (normal fasting glucose levels are less than 100, but we don’t yet know what level is normal after eating).

Except anyone with a glucose meter...?

The fact the author, a cardiologist, was surprised that oatmeal spiked his glucose is concerning to say the least, and certainly speaks to the eduction doctors get on nutrition.

I'd highly recommend self experimenting for a week or two, to establish 1. Fasted glucose 2. Fed glucose on your normal meals.

This is big in the keto space, with a lot of data being collected and shared, but apparently beyond the scope of the Times.

The meter I, and most people use: https://www.amazon.com/Precision-Glucose-Monitoring-Freestyl...

I don't quite understand why the microbiome should matter so much. Shouldn't it adapt to your diet if you stick to it long enough?

Bacterial persistence is weird. There's a Radiolab episode[1] where Robert Krulwich and Neil deGrasse Tyson shake hands and then measure the bacterial cultures on their hands repeatedly. Both shared bacteria, but the half-life of Tyson's bacteria on Krulwich's hands was short, while Krulwich's bacteria never seemed to disappear from Tyson's hands. (I don't remember how long they took measurements for, though, and I don't think it was very long.)

[1] https://www.wnycstudios.org/story/handshake-experiment

This trend of micro managing and optimizing every single details of our lives is very weird to me, some people seems to spend more time optimizing than living.

On one hand you'll get your DNA sampled and an "AI" to tell you what to eat, on the other hand we spend 10+ hours a day sitting, live in heavily polluted cities (light, noise, air), don't do anywhere close to enough physical activity. If you eat what's considered a healthy diet [0] these things will impact your life quality much more.

What people need is self control and education, we use food as a source of pleasure instead of seeing it as fuel (this could be a symptoms of an unhealthy lifestyle in general, but that's another topic). People eat 3 times a days + regular snack even if they are not hungry. We lost our ability to listen to our bodies, but luckily it's not very hard to learn it back. No one needs an AI to know that a bottle of coke a day, a bag of m&m's or a 500g cup of non dairy, gluten free, #FeelGood, #BodyPositity cup of ben& jerry's a has a net negative impact on your body.

The whole system is fucked from top to bottom, look at supermarkets, half of the aisles are full of colorful, family sized bags of candies, chocolate, ice creams, frozen junk and sodas with cute little mascots to attract kids. Society lost its way so bad that we now celebrate obesity and ill health. The obesity epidemic is an insult to humanity.

Sure, a tiny minority of people have very specific needs due to rare diseases that could benefits from these kind of studies. But for the rest of us avoiding processed food and not over eating is more than enough. If you gain a kilo every 3 months and have a sugar rush after every meal your brain is more than capable to know what's happening, it's a matter of taking personal responsibilities and facing your problems.

> specific food recommendations in order to avoid glucose spikes

Healthy people (aka the vast majority people) don't need any "AI" or "machine learning" for that, just basic nutrition education and a pair of eyes to look at the nutritional facts of the products they're about to buy.

There is a step after which the returns are greatly diminished, and for the vast majority of people that step is way closer to "pay attention and educate yourself" than "do what this AI tells you to".

> Using machine learning, a subtype of artificial intelligence, ... . In that way, an algorithm was built without the biases of the scientists.

I'd like to know how they determine that. "AI" and "machine learning" aren't unbiased by nature: https://www.newscientist.com/article/2166207-discriminating-...

"We are endowed by Nature with an interest in our own well-being; but this very interest, when overindulged, becomes a vice. Nature has intermingled pleasure with necessary things – not in order that we should seek pleasure, but in order that the addition of pleasure may make the indispensable means of existence attractive to our eyes." - Seneca

[0] we all know what a healthy meal is, let's not pretend it's a secret, eating healthy doesn't mean hitting all your blood test metrics with a nanomole precision.

Yeah. It's kind of embarrassing as a human to see this stuff crop up over and over.

If you walk into a supermarket, ignore the processed food, and essentially eat as much as you want of everything else, you'll be mostly alright.

Theoretically someone could gorge themselves on a million avocados or solely eat pork fat or something but they're not going to do that.

Almost all of this stuff comes about because people treat non-food as food.

Can you back up those assertions with some sources? This seems overly simplistic.


If you need a source to tell you that eating bullshit food is a problem, I suggest you re-read the parent post again.

You're letting some sort of pursuit of perfection blind you to obvious facts in front of your face.

> What people need is self control and education [...] People eat 3 times a days + regular snack even if they are not hungry.

Don't underestimate the effects of addiction, spurred on by gut bacteria manipulating your cravings. It is much less a matter of knowledge and will power than you think, it is very difficult to defeat your own brain.

Having worked in the health and fitness space for half a decade, this is so incredibly true. I was a chubby and inactive kid who discovered weight lifting and nutrition thanks to some helpful forums. I developed the habits, routines and discipline to get in shape, and maintain it for decades. In so doing, I thought since I could do it (the chubby non-athletic kid), anyone can. And should.

Yes, reduce your calories and you will lose weight. But that is missing a vital part of the causality chain. A nuked gut biome, thanks to your overworked and underpaid parents feeding you microwaved carbs in sugar sauce every night, is very hard to overcome. As is decades of government propaganda telling you to eat the exact opposite of what you should be.

It sure is an addiction, and as theNJR said it has multiple factors / causes.

But starting the wheel and wanting to better yourself is the first step and this one is purely a personal one, we now have the tools and the knowledge to fix these issues, what we lack is the will. Taking care of the fact that society screams "C O N S U M E" at every opportunity is another issue that should be tackled.

We're in the so called 'information age", we have more knowledge at our finger tips than the elite thinkers and scientists had in the last millenniums, yet people never ate so bad in recent history. I'd take my great grandparents deficiency in some vitamins over obesity or t2 diabetes any day.

The trend of optimising every detail of our lives is kind of inevitable. This realisation is what created the transhumanism philosophy.

Natural selection and evolution doesn't just happen at the biological level, it also happens on the informational and decision making level. If you're not optimising all your life decisions, you cede to someone who does.

> If you're not optimising all your life decisions, you cede to someone who does.

You're assuming that such optimization is done with a good fitness function. If the quality of the fitness measurement is poor, it doesn't matter how hard you try to micro-optimize, you'll be no better off than someone who merely puts some care.

Are you sure we don’t know what normal blood glucose ranges are after eating? I’m pretty sure they are between 120-140 as an upper limit. I also thought post-meal BG was exactly how we diagnosed diabetes? Clarification on that point would be helpful.



I can understand that absolute values aren't very meaningful, as they're probably missing a lot of necessary context. BMI is one metric that doesn't paint a clear picture on its own -- but no doctor would ever rely solely on it.

I can't understand dismissing most of medical science. Can you expand on that part of your statement?

Dr. Topol is making some reasonable observations and then extrapolating wild and wholly unsupported conclusions from them. The Mediterranean diet data is better than many pharmaceutically funded drug studies that medicine bases wide ranging guidelines on. It’s harmful for scientists to be led into such wild speculation when there are plenty of studies with endpoints like death that are measurable and real vs these bizarre end point measures.

How does the algo measure success? It it running analysis on my human, uh, “outputs”? If so then, sure, maybe.

This algo was measuring after meal glucose levels and optimizing against glucose spikes. Read the article before asking questions about the article.

Oh man, now you can pay to have AI tell you not to eat high-glycemic, insulin spiking foods. Wait until people figure out the algo wants them to eat less carbs ... "Terminator ... Skynet ... The Matrix"

God forbid it ever mention "keto"

The magic here is that some foods which we would expect to spike blood glucose don't for some people, whereas others that we would expect to keep blood glucose stable in fact spike it. It's making the case that people's blood glucose response is in fact highly individual, due to our highly individual mix of gut flora.

microbiome is ephemeral & adaptive.

The first attempt at this ended in failure.


Complaining about the philosophy aside, I'm trying to figure out the tech here. How was blood glucose monitored? A continuous blood monitor, or something over the skin?

Most likely a Dexcom G6.

But if you are interested in testing yourself, get this: https://www.amazon.com/Precision-Glucose-Monitoring-Freestyl...

You extract a pin drop of blood and it shows your glucose in a few seconds.

no thank you

"Siri of thousands of years from now, exactly how did the conditions which eventually prompted the Butlerian Jyhad begin?"

No thanks I don’t want to give away my last piece of freedom.

I don't think people use these to deny themselves. Rather, they use these to achieve goals they've otherwise had a difficult time achieving. In some way they gain freedom - freedom to achieve important long-term health goals, and freedom from short-term appetites.

Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact