All of bvbvbvbvbvbvbvbvbvbvbv's Comments + Replies

Why can't jurors vote for themselves if there is no maximum limit to the number of vote?

I'm interested in knowing your opinion on one of my ideas:

Instead of selecting 16 000 people at random and give them a vote. Select 16 000 people at random and give them a vote that can be passed up to 3 times to another person the former finds more competent/smart/informed.

I can totally imagine competent and smart people being sad of being selected at random while having not enough time to handle it (what if you just had a baby? etc) and would much prefer one of their trusted acquaintance do that.

The expected consequence would be that instead of having 16... (read more)

Things you might consider :

  • eating capsule of complement fibers when drinking Huel. This slows down the absorption process and might reduce the spikes and comedown.
  • add a tasteless oil (canola oil for example) to the Huel after adding the water. Stir a lot to make the two mix well (if needed add an emulsifier like soy lecithin or use a milk frother). The end result will be just like Huel but with a creamy texture and A LOT more calories.
  • Some declination of powdered meals have extra proteins and less carbs. That might be a way to reduce the spiking.
  • hard boil a batch of eggs in advance and eat them every so often.

Careful not to slip into ketosis without meaning to.

Thanks a lot! Would you mind pointing me to a direction that allows me to stay up to date on this niche? Also, do you have any contextual information to share about the paper you linked? Was it controversial for example?


2Garrett Baker2mo
It wasn’t controversial. People have theories about consciousness and neural architectures, and they use simulations to verify their theories produce predictions consistent with the real world. For more, you can probably follow the authors of this paper on twitter or something, or look through their backlogs, and do the same for Stanislas Dehaene.

You talk about the length of your sleep but have you considered variability of your sleep? For me sleeping always 7h30 seems to greatly increase my productivity in the long run. I sometimes add a few sleep cycle every few week ends but try to stick to a regular sleep schedule instead of thinking in terms of duration.

You might want to try by first waking up always at the same time, regardless of when you fell asleep.

A one word answer here is not very useful I think but I somewhat agree.

Adderall and other ADHD medication can, in a context of abuse, lead to somewhat "more productive time" or something like that. This if well known.

But too few people know that having a reduced "attention store" throughout the day can be a symptom of ADHD.

I can recall a patient that was not overly hyperactive but was daydreaming a lot and starting the medication actually helped him go "slower" in his mind but for way longer, hence increasing output.

Adderall Risks: Much More Than You Wanted To Know []




I'm a bit late to the party. I postponed writing this because it felt so important that I wanted to reply correctly. I sincerely hope that my enthusiasm will not be mistaken for egotism. Forgive my Chutzpah!

Let me introduce myself. After a background in hard sciences I am currently a medical student in a european country (hint, president's name sounds like a pastry and loves nuclear energy). I am confident that I will become a psychiatrist (i.e. I am still a few years before the final exam where you choose the specialty but it's pretty easy... (read more)

The implementation is probably very inefficient and I suffered from my own scrope creepiness but the idea is still there.

Posting this comment motivated me to fix some bugs in the next few days so if you run into issues give it another go next week btw

To keep motivated on my goals I coded my own tool called LiTOY (list that outlives you). It's FOSS in python :

I'm the only one to use it so there are probably bugs etc but maybe reading the can inspire you :)

There is no functionnal GUI so it's in CLI but I plan to restart it from scratch in typescript as an addon for logseq

ping @JayMon who was interested by it 2 years ago (time flies)

Nice. You might have solved the problem (will see if I actually like using it) where making a tool like this was on my list to prioritize my list, but I never got to it because the system did not exist.

Currently being a medical student that's very into AI, a dream of mine is to be in independant researcher in computational psychiatry.

Your post is very inspiring.


I am not aware of any way to quantify (even naively like my system) this kind of thing and I am very eager to hear about other ways people have found.

You described using it for 'bubble evaluation'. I've also heard of stuff like that to measure bias. Which thing, and what kind of thing?

A metric for comparing Social Circles

Epistemic status : Just an idea I had on a walk, doesn't seem that stupid to me

I have been thinking a bit about this topic lately, had an idea of a solution and figured LW would be interested in pointing out the unavoidable flaws in the reasoning.

Here's the gist Find a formula to quantify, as objectively as possible, your filter bubble (also called social bubble or even social circle). One could also see this as measuring by how much your social circle differ from random. The metric I chose to focus on is the income in... (read more)

I wouldn't say there's flaws in reasoning. Just that multiple comparisons are more likely to have issues, it's just a proxy, etc. It's an interesting idea.

My several cents :

It's not just about celiac disease. There are several other disorders that seem to be worsened by gluten, the first one that comes to my mind is endometriosis.

The way I see it, it seems that some people are susceptible to pro inflammatory effects of gluten, whereas most are not. So if you already have an inflammatory disorder or autoimmune condition AND have this inflammatory sensitivity then gluten will worsen it.

I don't know if this susceptibility to gluten's inflammatory effect is genetic, or genetically identified / identifiable. Whereas you can indeed identify celiac related antibodies using various tests.

2G Gordon Worley III1y
This fits my model that gluten somehow contributes to autoimmune/inflammation "load" and that because I'm now dealing with chronically worse asthma, even with treatment it may not be getting me down to a low enough level to consume as much gluten (or something else!) as I could in the past without issue.

I don't think this question has been asked before here but I'm wondering about the effect of the age of the dog at the start of the training.

  1. IIRC feral children that learn to communicate as adults have missed their critical period for grammatical expression. In effect there oral skills are akin to "dog speak". Maybe if the dog starts at age 0 it can lead to more complex word associations. If indeed there is a "grammatical critical period" for dogs even though dogs don't ever have to use grammars in their natural state that would be a breakthrough in my

... (read more)

Regarding the fact that you forget quicker than usual. Why not using the deck settings like lowering the interval modifier + adding a 6 month limit ? Or just lowering the starting ease.

Because I'm lazy. More generally, usually, the path is: I see someone every week, then I don't see them at all anymore. This is usually the case when I move to a new country/job. That's the moment where the interval/setting should change. Even if I don't do it manually, the next time the question is asked and I press again, the fact that I forgot means the interval will become small again. In this case, the app will mostly deal with it by itself, even if it takes some time, so no reason to take time dealing with it myself.

It will have been a year in a month and a half. We are currently at 1.33 million deaths. We are not going to have 3.7 million deaths in the next month. For why that won't happen regardless of the amount of policy attention see

I don't follow. I know there won't be 3.7M deaths in a month and a half, I'm arguing that without policies we would be at more than 1.33 million deaths and would plausibly end the year at more than 5M. If you disagree, could you point me to an article instead of the whole zvi website? I have no idea w... (read more)

  1. You could add the added uncertainty. Covid had all the more reasons to gather lots of attention at its beginning because we had no idea of the possible death rate. Whereas tobacco's risk has been known for a while. We still don't know anything about long term consequences of such an infection. Maybe none, maybe not.

  2. I think most people consider than smokers more or less choose to smoke, whereas covid kills and cripples far more arbitrarilly. This makes it way more of a threat for most people, who "could just decide not to smoke".

Aside from that, can... (read more)

It will have been a year in a month and a half. We are currently at 1.33 million deaths. We are not going to have 3.7 million deaths in the next month. For why that won't happen regardless of the amount of policy attention see []. We know COVID has a barrier to reinfection, so Covid is very unlikely to "circle the world for years". Also the tobacco deaths are actually going to continue for decades, so this can't be an argument for more marginal attention to Covid. Do you believe the marginal cost of preventing a Covid death is lower than the marginal cost of preventing a tobacco death? Why or why not?

The real problem is assigning values to the different tasks. How valuable is planning out my course schedule for the next year? Evaluating a specific research agenda? Exploring the papers in a new field? Unfortunately, I lack good answers for these questions. I'm vaguely aware that I should have >5 articles published when I enter the job market. But the weightings for quality, quantity, prestige and coherence of these publications are unclear. This is a vital area for further research.

I have been slowly developping my own python app to solve this. I ... (read more)

What are your opinions on STAR vs majority judgement?

To me, when I was reading about the different voting systems at the time of the last election in my country I recall majority judgement was a very clear winner. So I'm surprised to not see it mentionned here.

Personnaly I don't. I just call some friends individually and organize a serie of 1-1 dinners.

The science youtuber behind the channel "the thought emporium" and "the taste emporium" has an interesting take on this : at 25 seconds. Basically he makes a cake outside of his skill set, spends the whole day on it and give it to his friends.

If he's sentence was 1 word away from right to super wrong I still think it's terribly unprofessional.

In any case, he should know that the sentence is absolutely critical. He should have made three setences in a row to clarify just in case, because if he missed the word "dying" from your sentence the consequences are enormous.

Disclaimer : I have not read studies on zinc and covid, and barely some with zinc and flu. My answer assumes that the consensus is that zinc shortens and reduces the severity of the common cold. Also that zinc and covid-19 are not studied

Traditionally, the medical field doesn't put much energy into studying what self care is effective. It's rather a field that does it best to outlaw self care.

I think "outlaw" is an exagerration, I would rather say "cautions against".

From what we know about Zinc lozenges and the flu, they have to be taken at symptoms

... (read more)

Good point. It was indeed not clear but I meant "known" not as in "by everyone" but rather more by its occurence in my readings, which are more technical and rigorous than mainstream media. (I am in the medical field)

Traditionally, the medical field doesn't put much energy into studying what self care is effective. It's rather a field that does it best to outlaw self care. When it comes to medical treatment for COVID-19, it's treatment that's effective for people who are diagnosed with COVID-19. To be diagnosed with COVID-19 you usually need to have symptoms and then wait one or two days for your test. That time allows the virus to replicate a lot and as a result treatments that might be useful to be taken a day before overt symptoms when HRV falls for an otherwise unexplained reason or at symptoms onset might not be useful in the clinical setting for diagnosed COVID-19 patients. From what we know about Zinc lozenges and the flu, they have to be taken at symptoms onset to be effective, so the assumption that the same is true for COVID-19 is not farfetched.

That's right, my reasonning was flawed. Thank you!

Disclaimer : I have not researched any guidance for self care, just sharing one reasonning here.

I have no knowledge about this whatsoever, but from a population standpoint I'd be surprised if there was anything that you could buy that will probably help you while not posing comparatively more risks to others by the simple act of going shopping while positive (thus spreading).

What I mean is that I think everyone in the area is better of if they isolate as soon as they know they're positive instead of first going shopping for something that might help.

Of co... (read more)

For shopping, the internet exists!
Given how long it took for the expected gain of wearing mask to be known it seems very optimistic.

It might be out of the scope of what you're asking but personnaly I'm very interested in the philosophical constraints of the models used in economics.

Specifically, I found that I can learn on pretty much any theoretical model of finance quite easily if I spend enough time on the internet.

But, in the process of reading about it, it's actually very hard to know if it takes for granted a specific philosophical model with specific restraints.

I lack a global understanding of the philosophy of economics and it worries me given the supposed hegemony of the US-bo... (read more)

It's going to look like it's coming from newhere but here's one opinion.

There is something strange when people take a large dose of some serotonin agonists : they think that the universe is wonderful and life has a purpose or something like that. They basically find purpose and emotion in something, without usually being able to rationnaly explain it afterwards.

Reading about someone who apparently had enough cortex to become an MD then somehow linked 5G to viruses makes me think : is there the possibility of a neurological [bug/change/disorder] where extreme purpose is derived from some random factual event?

To me it could potentially be the same system. The "purpose finder" system.

Glad I got my point across, I was really not sure it was clear.

My position is that this matter is complicated and it looks hard to gather more data. For example I can't see a satisfying way to make surveys about what those people think. Especially : the "conspiracy theorists" that are actually prone to doubt could be so susceptible to this bias they if you ask them about the president being a lizard they would answer of course without actually believing it prior to the survey. (Maybe something to do with a very short term attention span coupled with a para... (read more)

I am not convinced. I feel there could be a huge variance in how people actually believe there conspiracies. And it look extremely hard to estimate it.

Of course I have no doubt that there are/will be people that actually believe these things. But it could be that the proportion of people who believe them is small while the proportion of people who don't believe it but are now doubting is large.

I have never been entirely convinced that it's implausible that there is a majority of the conspiracy theorists that are actually susceptible to a cognitive bias tha... (read more)

Wow, you actually made me a bit optimistic about this! Indeed, if foreign propaganda makes things worse, that means the domestic conspiracy theorists are actually less crazy than they seem.

You are completely right. I should have known better. Allow me to downvote myself.

I found the first question really interesting. It would be interesting to make it recurrent (say every week) and see how quickly equilibrium would be found and how stable it is.

I must point out that I think it was not well phrased though, I thought it meant something like "pick the subject most readers will not choose to read" or something like that. This could explain why the distribution is so far from 25% each.

I think the distribution range is roughly what you would expect by chance given the number of responses (when I did it: 65 responses total, range of results 18.5% - 30.8%)

It will be interesting to see if the antivax movement dimishes its spread if COVID-19 ends up being eradicated from countries where vaccination campaign have been led, and is still killing left and right in countries where antivax and distrust of science is common.

Especially, it would mean that the cost of antivaxxers will be enormous to the country. Maybe this would incentivize government to act more on it.

Do not underestimate the creativity of conspiracy theorists. Allow me to give you a summary of Russia-sponsored conspiracy broadcast in my country: Stuff like this circulates on social networks, along with warnings not to give children face masks, because face masks cause infections (which is exactly what Big Pharma wants, so that they can sell you the cure later). What actually helps is vitamins C and D, and most importantly, positive thinking and avoiding negative messages from mainstream media!, I don't really see how numbers of dead people could convince anyone. You can always attribute them to a different cause. And on the English-speaking web, people still celebrate Swedish "success", don't they?

I think that you would have to only consider countries that have similar suicide rate as well as conviction rate, as I've heard that some countries (Japan) rule every unsolved homicide as suicide to have a perfect police track record.

I am not qualified to estimate if it's true but that's something else to consider.

Indeed I was talking about "Deep-learning based AGI". Thank you for the thorough answer.

For one, I don't think it's reasonnable to assume that the future GPT-N will only work in a "text continuation setup".

But also, what would happen if you had it read a database of exploits and ask him to evade the API, or to "create a new exploit to break this API" or something like that.

I don't work in the field so it's a question.

future versions of such models could well work in other ways than text continuation, but this would require new ideas not present in the current way these models are trained, which is literally by trying to maximise the probability they assign to the true next word in the dataset. I think the abstraction of "GPT-N" is useful if it refers to a simply scaled up version of GPT-3, no clever additional tricks, no new paradigms, just the same thing with more parameters and more data, if you don't assume this then "GPT-N" is no more specific than "Deep Learning-based AGI", and we must then only talk in very general terms. Regarding the exploits, you need to massage your question in a way that GPT-N predicts that its answer is the most likely thing that a human would write after your question. Over the whole internet, most of the time when someone asks someone else to answer a really hard question, the human who writes the text immediatly after that question will either a) be wrong or b) avoid the question. GPT-N isn't trying to be right, to it, avoiding your question or being wrong is perfectly fine, because that's what it was trained to output after hard questions. To generate such an exploit, you need to convince GPT-N that the text it is being shown is actually coming from really competent humans, so you might try to frame your question as the beginning of a computer science paper, maybe written far in the future, and which has lots of citations, written by a collaboration of people GPT-N knows are competent. But then GPT-N might predict that those humans would not publish such a dangerous exploit, so it would yet again evade you. After a bit of trial and error, you might well corner GPT-N into producing what you want, but it will not be easy.

Computer security is an endless game of cats and mouse. Here you showed a URL pointing to something the cat knows. But there are plenty of literature/db on what the mouse have learned throughout the years.

If an AI became somehow self aware and had access to the knowledge of all the previous mice and cats, I wouldn't be surprised if it could break free. But that's a big if.

I agree with the cat and mouse metaphor and that we should assume an AI to be hyper competent. At the same time, it will be restricted to operating within the constraints of the systems in can influence. My main point, which I admit was poorly made, is that cross site scripting attacks can be covered with a small investment, which eliminates clever java script as a possible attack vector. I would place lower probability on this being the way an AI escapes. I would place higher probability on an AI exploiting a memory buffering type error similar to the one you referenced. Furthermore I would expect it to be in software it is running on top of and can easily experiment/iterate on. (OS, container, whatever) Whereas browser interactions are limited in iteration by the number of times a user calls the service, one would expect the local software can by manipulated and experimented with constantly and only be constrained by the CPU /IO resources available.

This makes me wonder, do AI researcher working on GPT-N corpus intentionnaly avoir putting their own work?

Does GPT have, in its corpus, details about the implementation of itself, or past attemps, or access to blogposts from people like you suggesting this very idea?

Or do they try to remove it from the corpus?

edit : A friend working in AI told me that they take enormous care of what they include and that, to the best of his knowledge it doesn't contain research paper. He doesn't know however if it includes wikipedia pages related to AI, or to the creator of itself etc.

Thank you very much for taking the time, I can't answer for the time being as much as I'd like so I won't. But I'm linking your comment to my todo of the project so I can get back to you when I'll have more time for coding.

But most of what you suggested is already planned, which is reassuring to me, even though your first draft is as advanced as my long reflections on it I must say. Have a nice day!

Very interesting!

Could you explain the workflow?

Also, do you intend to make the code accessible?

Funny you're asking this, I have the same problem and had an idea about a solution.

Even though recent events forced me to delay the coding, I am hopeful that this can help me do a dynamic semi-selfsorted list where I can dump all the things I want to read (be it articles, books, studies etc) and help me get to the bottom of it. I will of course make the code available somewhere.

Here's the link to my previous comment that explained this

Here's the text directly :

The idea is to have an sql database that contains everything you want to do. The python script

... (read more)
Interesting idea, the only thing I can think of adding is maybe add tagging to the system. In the case of your media: eg, if you feel like watching Sci-Fi you can run a search to retrieve some of the highest recommendations from your backlog. Of those you may have results that come back with anime, Western, and foreign live-action results. If there's a specific type of show you want you could then add that tag to your results (anime+sci-fi) to get a more precise list of recommendations. Tagging could be both top-level (as above, picking Sci-Fi doesn't infer that it must be anime) or hierarchical. I could think of something like programming. You have a set of different programming projects that you would like to work on, however they may require different libraries or languages to work (golang, jscript, R; or Tensorflow, Node.js). You could tag them as resources necessary to complete high-level projects. Using an ELO-scoring system you could compare projects to get a rough estimate of what should be done, followed by running similar scoring on the child tags to determine either the order to read, practice skills or tasks to complete. If your work tasks are separate from your read tasks, then this might offer an extra benefit by showing you articles/reference material that is applicable to (estimated) upcoming or current projects. Top-level tags might add a constant increase to ELO scores of subtags, however if some subtags start to outperform other subtags in differing parent tags, the system might reorganize your projects to highlight ones you would prefer to focus on. If it keeps track of tasks done it may also force you to handle tasks that while low ELO by themselves are necessary to complete a nearly finished project. One problem might be chaining tasks, where (an estimated) preferred task exists behind a less desirable task. That might cause a hit to ELO even if it would be better in the long-run to finish the detestable task. Maybe using the higher ELO task

Studies that seem to show that the brain cannot estimate the loss of his abilities when it's missing sleep. Which has great implications to night driving for example and is definitely not intuitive.

I am a student in a field completely unrelated to any of this and I tend to have my hobbies take too much of my time compared to what my uni asks of me.

The consequence is that I try to optimize the time I spend here. So I add what feels like future high value in my diy reading queue. De facto almost anything that touches to AI that doesn't seem too easy ends up in this queue. Then what I think will be useful to me (eg : the "how to be happy" essay that I discovered thanks to your comment above).

But I never add to my queue (or ignore) before reading it diag... (read more)

I am mostly interested in the meta. By that I mean how people get their points across or how do they developp their ideas etc.

LW is full of very interesting writing styles and approaches to questions. It interests me usually more than the subject of the question.

Otherwise : anything related to AI or how to think ends up in my reading queue.

Thanks for the answer! It is pretty interesting: do you really not care at all about the ideas themselves (except the two topics mentioned)? A related question might be "how do you decide to go read a post only from the title, if you only use the meta?"

Funny you mention this as, starting tomorow, I am writing a script that tries to solve this issue.

The idea is to have an sql database that contains everything you want to do. The python script picks 2 entries and ask "If you die in a year from now, wich activity minimized regret the most?". The answer will probably be a cursor from activity1 <-> neutral <-> activity2. The scores are then calculated using the ELO scoring. This pair comparison algorithm allows to quickly rank your lifegoals by importance.

But that's only half of it.... (read more)

2Evan Ward2y
It's great to see other people thinking and working on these ideas of efficiently eliciting preferences and very 'subjective' data, and building your own long-term decision support system! I've been pretty frustrated by the seeming lack of tooling for this. Inspired partially by Gwern's Resorter as well, I've started experimenting with my own version, except my goal is to end up with random variables for cardinal utilities (at least across various metrics), and I'm having the inputs for comparisons be quickly-drawn probability distributions.