Hacker News new | past | comments | ask | show | jobs | submit login
Death Note: L, Anonymity and Eluding Entropy (2017) (gwern.net)
525 points by tosh on Aug 5, 2019 | hide | past | favorite | 139 comments

I would frame things a little differently. Light makes a lot of mistakes because the Death note is incredibly powerful and there would be almost no reasonable way for L to catch Light without Light making obvious mistakes (and would also make for a boring story). Undoubtedly, the biggest mistake Light makes is revealing the existence of the Death Note, and all the rules therein, which not only gives L (almost) all the information he needs, but also traps him from any subsequent activity.

Without knowing the rules of the Death Note, I don't think L could have done anything (besides outright abducting/murdering Light) to stop Light, even if he is 100% sure Light is "responsible".

That said, I feel like a much simpler solution than what's proposed at the end of the essay is:

1. Gather up the names you want to write over the course of a week. 2. Wait another week, then write it at a specific point of time of the week. 3. Keep to that schedule.

The names need to be as public and as well known as possible such that it would be impossible to trace the flow of information. (Remember, the goal is fear, not specificity.) Keeping to that schedule means you will never leak more information thereafter.

I agree with this completely; Light makes up for what I would say is an inferior deductive ability by using a supernatural item with rules that defy ordinary logic and reason (L catches on early that such a tool must exist with his TV announcement stunt, but he can't possibly know the ruleset of such an item).

The only way Near and Mello end up overcoming this advantage is by: (1) Mello obtaining a Death Note of his own, allowing him to ascertain the real ruleset, and (2) Near not being directly involved in Light's life the way L was, and there not leaking any extra information about himself until the finale. That last bit is especially critical considering L was taken out precisely because he was physically present and therefore had his real name observed by Shinigami eyes.

I think the most under-utilized ability of the Death Note in the series is the ability to specify manner of death. Light uses this on occasion, but he could've have made many criminal deaths seem accidental or suicidal in addition to staggering their occurrences in time in order to throw L off from the very beginning.

>I think the most under-utilized ability of the Death Note in the series is the ability to specify manner of death.

Absolutely. This ability is so significant that it completely overshadows the ability to kill people. If anything, the deaths should be treated as an unfortunate byproduct of being able to control people for 23 days in a row.

There were limits to this ability though. All behaviour had to be in character.

Agreed, but there is still so much you can do. You can make dictators liberalize their countries. You can make billionaires donate away their wealth. And so on.

As far as I remember the only reason Light lost at the end is because a character managed to make an exact replica of a Death Note overnight. A magic book that can hold infinite names. It was this plothole that ruined the ending for me.

To be fair to him, it's generally portrayed as a generic black notebook with words on the front, and the only way to notice that lack of infinite pages is to write hundreds or even thousands of names.

I was more concerned with the ability to copy infinite or even just a couple of hundred sites overnight.

> Remember, the goal is fear, not specificity.

I think that's where your proposal breaks down. Light is powerhungry and arrogant, but he's also very moralistic. A character that proritizes fear over accuracy would be a different character than Light. He wants to build a better, even perfect, world. And for that, he must have accuracy. Otherwise he's a mere tyrant, something he believes he is not.

Of course, it's worth saying, the great archetypal tragedy of the story, and what makes it excellent literature, is that his goal is unacheivable. Humans are deeply imperfect and powerless to end evil completely. And the process of trying turns us into the worst of all monsters that can possibly exist.

>Light is powerhungry and arrogant, but he's also very moralistic. A character that proritizes fear over accuracy would be a different character than Light. He wants to build a better, even perfect, world. And for that, he must have accuracy. Otherwise he's a mere tyrant, something he believes he is not.

Somewhat. You're right that Light is extremely ambitious in his goal, but fear is also explicitly his chosen strategy. (Also, Light very clearly leans into being a tyrant. He says that eventually, he wants to start executing useless people as well.)

Another thing to consider is that (and I believe this happens in the manga), the news reporting itself responds to the Kira phenomenon, and starts publishing names accordingly. For us, let's think "Top 10 people in the world Kira should kill right now" listicles. The upside is that the information becomes more readily available. The downside is that you (our hypothetical, less-evil Light) may eventually want to exercise some discretion over who's being killed, and that would be where information gets leaked, and that's where I think this strategy might falter.

Is that really meant as the tragedy? I'm not sure to what extent it's really intended that way. Light's attitude to criminal justice is a distressingly realistic one in a country where 99% of arrests lead to confessions.

Just a little light reading that's close at hand from a few years back: https://www.economist.com/leaders/2015/12/05/forced-to-confe... https://www.economist.com/asia/2015/12/03/silent-screams

See also any critical commentary on Nissan and the extraordinary case of Carlos Ghosn.

What happened with Ghosn?

I can't speak to the intent. Upton Sinclair's Jungle was a socialist manifesto that accidentally missed America's heart and hit her gut instead.

What I can say is approximately 100% of material that gets popular is simply the retelling of something like 7 archetypal stories that our human minds resonate with. Watch Moana and Wreck it Ralph, they're the same story: the Resurrection of the Spirit of the Father. Watch Harry Potter and LOTR and the 5th Element, they're the same story as David and Goliath: the Overcoming the Impossible. Death Note and Faust are the same story: the Pact with the Devil.

I could go on.

I always hated this sort of statement because it confuses story with theme. Sure Harry Potter and LOTR and the 5th Element all might have some superficial similarities, but each of them is very different in ways aside from mere set dressing.

You can call anything equivalent if you pick and chose what factors matter

Indeed. It's always "just" or "simply" the same <n> basic stories. But it's hardly simple to go from the 3 words "overcoming the impossible" to Lord of the Rings.

Specifically, the power of the <n> basic stories meme is that it implies a lower information content of our stories than we supposedly realize. The meme makes you feel smart by inducting you into the enlightened set. But in fact, the low information content is in the supposed categories, not our stories. They throw out all the details that actually make a story. Even if one of the categories was the ridiculously specific "tiny man makes treacherous journey to destroy dangerous magical object, also there's a wizard", that's still a far cry from reconstituting LotR.

Hang on. I think you're missing the point. It's not about how "if you've read one, you've read them all". The fact that MacBeth is another way to tell the Futile Struggle Against Fate story doesn't in any way diminish the creativity or power or uniqueness of the work. The point is that if it's not an archetypical story it won't resonate with the human mind, and the writer will fail.

We tell the same stories again and again because there's a finite set of stories that engage our mind, whether we like it or not.

But my point is the archetypes are so broad they may as well be meaningless. "Futile Struggle Against Fate" encompasses so much.

May as well say all stories are "Man vs Man" or "Man vs Nature" or "Man vs God". Sure you can categorize basically every story into one of those three things but it is almost never useful to do so

It's not about hiding. Light was aiming to rule the world, as in completely eliminate any opposing parties, and you can't kill them without revealing part of yourself.

Death Note ended when L died.

This is a significant spoiler for those who haven't seen the show. Please delete it.

Do spoilers for a decade-old comic book/tv show really matter?

Yes, why wouldn't they? As long as it's culturally relevant, people benefit from experiencing it without spoilers. You shouldn't take that possibility away from them just because you already got yours.

I agree, it's not only a spoiler but adds nothing to the main subject under discussion.

lol wow and people were agreeing with me for a few hours there too but looks like consensus flipped! oh well too late for me to do anything about this decade plus old spoiler

@dang do your thang

everyone agrees with your point, but the fact remains that your comment was not only a spoiler but also added nothing useful to the discussion in this thread or to the thesis of the article.

I also didn't enjoy Death Note after this event.

This is somewhat related to a thought experiment of mine.

Suppose you had a means of transmitting information back in time - a one-node network that effectively could email to itself with a timestamp (in the past) being the message's "recipient."

Suppose such a thing existed in the world today. Assume the individuals in possession of such a thing were very intelligent, very careful and very guarded about its use.

Could you still detect that such a thing existed by observing real-world events - stock market changes, tech innovations, amazing coincidences and the like? What evidence would you look for?

If you possessed such a device then errors in your "predictions" would be indicative of another party using a similar device. As such, I assume multiple parties would soon discover each other, engage in an incomprehensibly complex time-battle, and there would suddenly (instantaneously in all patched timelines) be only one such party.

Furthermore, I imagine it would be very easy for such a party to accidently blunder into a situation where they lose control of the time-messaging device, erasing all evidence of its own existence and leaving the timeline perturbation-free until such "time" as reverse time travel is rediscovered.

In the book Singularity Sky [0], mankind invented the first time-looped computer, and it cycled up to a godlike overlord. After making a lot of very disruptive changes (which of course seem to be immediate to everyone else) it seems to have become surprisingly disinterested or uninvolved.

Anyway, it has only one hard "commandment": Nobody can violate causality anywhere inside its own historic light-cone, since doing so could be a potential attack (or competition) to its own existence.

[0] https://en.wikipedia.org/wiki/Singularity_Sky

Book was so much better than accelerando

Future Diary has some pretty cool time battles. While it's not as good as death note, it definitely has it's moments. https://m.imdb.com/title/tt2069441/?ref_=m_nv_sr_1

As a fan of time travel movies, I'd watch this.

You may like the anime Steins;Gate then. The plot is based on the idea of sending a small amount of information back in time and changing the sender's present, and then finds that another organization has been using this and is trying to control the world with the power.

Everyone in this thread should try Primer! Great little movie about a similar concept (smaller in scale though). Tries really hard to stick to its own rules.

Or an episodic series where every episode has a new main character inventing time travel and then accidentally destroying the conditions that allowed then to invent time travel.

Sounds like Dark.

It's approximately the plot of Primer

Spoiler territory for Evan Currie's Hayden War Cycle Series ahead - which would deserve some spotlight and an editor for catching all the spelling mistakes both...


This is touched upon by a favorite sci-fi story of mine, where humans come in possession of such technology and make huge leaps in capabilities in short timeframes. Those leaps are unexplainable to most alien species they are in contact with.

Especially later on in the books a huge part of the story is told from alien POVs. The author has some interesting ideas about how other species may find us impressive, and in what respects they may consider us primitive.

The books are mostly about on-planet fighting and some ship-to-ship battles though. In fact there's little politics and communication at all until the cold-war arc. So don't expect the books to dive deeply into it at all.

But I really recommend the books just for how many interesting ideas Evan Currie managed to compress into a fast-paced action story. They left me with a lot of stuff to mull over long after putting the books down.

Also recoomend Evan currie

It depends dramatically on two things:

1. the type of time travel in effect. I'll assume Marvel-flavor time travel, where going back is really jumping to a different universe that shares the same history till the arrival time. But if it's something more complicated then relaying info back might Marty McFade the prognosticator out of existence.

2. the level of information available to the prognosticator (future communicator). Is it near-total, so they could tell the recipient to "duck… now!" or is it just public info like stock prices, weather, and what's on the news?

Best case for the recipient is that for each change they make, they still have an ally relaying valuable info from the future that includes those changes. And also that said ally has near-total information to help protect the recipient.

Worst case for the recipient is that after they make one change, they are now in a timeline where their future ally no longer communicates with them. This is also the most detectable for others in the recipient's time because the recipient will likely make some big bets based on the one-shot info they receive, and won't have additional info to help avoid detection.

To continue the thought experiment: if you developed that technology in the year 2120, how would you use it, and would that be detectable in the year 2020?

Sounds like a different anime to me. El Psy Kongroo.

Steins Gate is probably the closest answer to this (in which the time travel mechanism can only send back information) and is my favorite time travel mechanics system. Timelines diverge into their infinite possibilities but in self consistent ways. That is, some timelines can only exist as a result of information received from other timeline’s future; meaning the past is dependent on the future of a different timeline which won’t play out, but has added a sort of immutable constant to the first timeline.

It depends. If these are spherical rationalists in a vacuum, they’ll know what you know about how to detect it’s use and set their usage to the highest undetectable level. Having a time machine might make this trivial to achieve, depending only on if it’s possible to change the past in this model of time travel.

Time machines can be used as computers to reduce PSPACE (which includes NP) problems to constant time.

Is polynomial space something to assume every time traveler has?

Even ones without a TARDIS?

PSPACE[0] is a computational complexity class, not about topology ("larger on the inside").

Any timemachine that produces self-consistent timelines (arrives at a fixed point after a finite amount of iterations) can be used to solve problems in PSPACE.[1]

[0] https://en.wikipedia.org/wiki/PSPACE [1] https://www.scottaaronson.com/papers/ctc.pdf

A particularly clever member of the Twitterati: https://twitter.com/Theophite/status/1062816091892674561

@Theophite: the existence of nonzero interest rates implies the nonexistence of time travel.

I see what they did there but I don't know if I agree. The market can stay irrational for longer than it takes to invent time travel...

The whole point of time travel is that it doesn't care about time; once time travel exists in the future, people in the future will be able to do all kinds of fun arbitrage with it.

Maybe, look for "Butterfly effects[1]" happening around you, i.e., small insignificant events that have huge, enormous after-effects.

Essentially, if there were a system that could send information from the future to the past, the future agent would want to be "as discreet as" possible, while still achieving "nearly perfect" outcome. Note that the future agent has the advantage of reiterating over and over his/her choices, because, well, the "future-future" he/she can still send a signal to his/her past about want went wrong the previous time.

This happens to be a common plot for a lot of various time-travel movies and TV shows like Dark, Umbrella Academy, etc.


There's an Asimov short story called The Red Queen's Race which deals with this question: what if the ancient Greeks had received a few hints about modern chemistry? I won't spoil how it relates more directly to your question.

Mark Twain's A Connecticut Yankee in King Arthur's Court

the original isekai?

This plot is the basis of another (very good) anime called Steins;Gate actually.

Assuming it's a reasonably high-bandwidth channel, very careful and guarded use seems much more boring than bootstrapping the singularity. So, hopefully you can find them and supplant them.

You very much would want to keep it guarded until you have secured your goal.

Bootstrapping the singularity is unlikely to be achieved with a single application of a time machine since you are only sending information to the past, not the technology itself.

And repeated application allows other actors to intervene during each iteration, which may result in the CTC settling into a different fixed point than the desired one.

You can't expect to own and use a time machine and be left alone while doing so.

You can write all of the advancements off as luck, inspiration, or drugs and blend in until some brain scanning technology is invented that could find out that you're being disingenuous with high confidence. By then, the idea of time travel might not be that exotic. Maybe you send back a way to fool that technology and then it's a cat/mouse game.

A fun read where the "could you still detect" portion turns out to be a very important plot point: Thrice Upon a Time by James P Hogan (https://www.goodreads.com/book/show/849488.Thrice_Upon_a_Tim...)

I would search for people living a "perfect" life: no mistakes ever made, not a single cent lost to investment, not a single car key forgotten or an umbrella staying at home, absolutely nothing negative in their life. The more there are, the less the chance that pure luck is driving it, it must be a communication from the future.

This seems too easy, if the thought experiment played out with anyone remotely intelligent I would bet they would purposefully make 'mistakes' so as to not gain too much attention from a perfect track record.

I wonder, can they A/B test alternatives with that time machine ? Would they settle for a comfortable 'small' one time lottery winnings? Or try and create a dynasty for a longer lived family/institution. Also the person in the future, would they 'join' their new history ?

If allowed to send a message back in time any mistakes made by such a person could be fixed presumably. It would seem to me that you would have to detect that someone is playing with time, find out who that is and then capture them before they learn that you know and send a message back to kill you.

Well, not if you make a mistake that leads to you being unable to send a message back. That's part of the plot of a sci-fi story I read: there's a machine that does this, and the messages stop at a specific point in the future. (It's 'the Arrows of time' by Greg Egan. The Orthogonal trilogy is good stuff.)

I’m not an accountant but I’ve heard that accountants often use number frequencies to look for fraud. IIRC, amounts of money starting with 9 are less likely. Looking at those kinds of distributions in the world can maybe show influence of non normal factors.

Assuming the owner is smart and does not want anyone to detect that such a thing exists, then no one can. As soon as the owner feels that someone else knows, they will send a message back as a warning and tell themselves to be more careful.

That is very nearly the plot of Primer

Very interesting article, my only real contention is on the point made in Mistake 4:

> You can see it was a gamble by considering if Light had been outside Kanto; since he would not see it live, he would not have reacted, and all L would learn is that his suspect was in that other 2/3 of the population, for a gain of only ~0.3 bits.

My understanding of L's plan (based on what he said in the broadcast) was that he intended to do several live broadcasts, one in each region, until he was caught. Lind L. Taylor was a death row criminal being offered a deal to survive, and would've gone along with doing several broadcasts. Additionally, even if they didn't do it live and simply recorded it there's no way that Light would've known it wasn't live and he would've acted in the same manner (and since Lind was in L's custody, L would've known if he died even if Light didn't).

At the time Death Note was written, mass internet communication wasn't as prevalent and so you wouldn't expect people to find out about events like the broadcast for at least a few hours (long enough to do broadcasts in each region). L probably hoped that Kira was in a less populated city since it would cut down on more bits of entropy (and I'd argue it would've been a better idea to start broadcasting to smaller cities first -- since it's more likely an event like this in a big city would be quickly broadcast to the rest of Japan).

From a pure efficiency point of view, if you are bisecting like this it is best to divide in halves, which gives you 1 bit of information each time regardless of the outcome.

Of course, this assumes you can continue to apply the same method many times, which probably doesn't apply to this case, as you can only fool Light with a similar method a limited number of times (possibly once only). In that case you probably want to do it more fine-grained than 1/3 so that you get more bits of information once you get to the first (and therefore last) successful attempt.

As a fun mental exercise, I've been wondering if the Death note could've been used to solve difficult computational problems. As far as I know, the following abides by Death Note rules, in particular we know that the Death Note is able to "manipulate" luck. Assume Bob Robert is an amateur mathematician, so this is something he would reasonably do and guess.

"Bob Robert. Writes down a number X>=82,589,933 where 2^X-1 is a prime number as his guess of the next largest known prime number, and it is indeed correct. Then dies of a heart attack."

If I remember correctly, it has to be something that the person was actually capable of, otherwise the person just dies.

The person is capable of writing down X, it's a handful of digits. Moreover it is not unreasonable for an amateur mathematician to do so.

The fact that they entirely lucky that the series of digits indeed satisfies the mathematical condition we want is incidental but I believe abides by Death Note rules.

I think "capable" also takes into account knowledge they have.

It's been a while since I watched Death Note, but isn't this brought up as a reason why Kira can't do, "writes down L's real name, then dies of a heart attack."?

It would be theoretically possible for someone to be lucky enough to guess L's real name, but apparently the note has a bullcrap detector or something where it eventually just says, "no, they wouldn't really know to do that."

Well we know the reason why Light says he can't do that is that "the person wouldn't know the name". But you're right in actuality that would allow for "person guesses the name, and is correct". And I think that is a loophole.

Let me put this another way. I think the following would be considered a valid Death Note command.

"Person flips a coin 100 times in a row, and gets heads every time".

Now how is that different from

"Person runs an RNG digit generator 100 times in a row, and gets 9s every time".


"Person runs an RNG digit generator 100 times in a row, the 100 digit number is exactly divisible by 123456".


"Person runs an RNG digit generator 100 times in a row, 2^x-1 of that number is prime".

And if this is all iffy, let me propose an alternative. Let's assume (I'm not 100% sure of this) that we can verify large prime within 23 days. Heck, build a data center to do this if we have to.

"Person runs RNG to generate x such that 2^x-1 is prime. Uses machine and verifies that it is indeed prime. Dies of joy."

You can similarly use this to crack hashes, etc.

The Death Note doesn't give the target any magic abilities, that the previously didn't have.

> "Person flips a coin 100 times in a row, and gets heads every time".

I'm pretty sure this would be interpreted as "The person has to flip coins until they have a 100 streak of only heads". This will obviously take a really long time, and the person might be driven insane before that. Similar things go for your other examples.

In a comment further up you wrote:

> in particular we know that the Death Note is able to "manipulate" luck

I don't recall that specifically. Even then, I would guess that this has it's bounds. I also have the feeling (not backed up by lore, but I think it would fit in the world), that bigger manipulations would take more energy from the respective shinigami, so while a shinigami might be able to manipulate luck enough to make someone win the lottery, finding prime numbers would be too draining.

>I'm pretty sure this would be interpreted as "The person has to flip coins until they have a 100 streak of only heads".

I disagree. For example, Light has people die of getting run over a by truck. But the Death note does not interpret this is as "person repeatedly crosses the street until they get run over by a truck". Instead, events are (reasonably) manipulated such that a truck runs them over. Similarly, the coins would be manipulated to have 100 heads in a row.

>I also have the feeling (not backed up by lore, but I think it would fit in the world), that bigger manipulations would take more energy from the respective shinigami, so while a shinigami might be able to manipulate luck enough to make someone win the lottery, finding prime numbers would be too draining.

I wouldn't be surprised if this is true, but hey, what's a couple of Shinigami to save us some compute!

My impression is that the Death Note would only manipulate luck to a certain degree -- ie, if the odds of even occurring are greater than 1/x, just make it happen. That would allow you to get hit by a truck crossing the street, but not allow you to win a coin flip 100 times (7.8886091e-31 odds). You'd probably have to experiment to figure out what the threshold was.

That being said... let's assume you're right. If the death note can manipulate odds to an arbitrary degree, you're thinking way too small right now. It would be much more efficient to write:

"X opens a laptop text editor, closes their eyes, and hits keys randomly. They open their eyes and discover they have accidentally written an algorithm for quickly factoring large numbers. They copy and paste this algorithm into an encrypted email to light.iskira@gmail.com, then die by jumping in front of a bus."

If X dies of a heart attack, you know your instructions didn't take and there's no algorithm for quickly factoring large numbers -- the problem was impossible. But then you can start modifying your query for, "an easy strategy for keeping cubits stable", or whatever you want to follow up with.

If I have a fast-track to get at almost any (possible) invention I can think of, I'm not going to waste my time factoring large numbers!

>If the death note can manipulate odds to an arbitrary degree, you're thinking way too small right now.

Right, what I feel might be the case is that the targeted output needs to at least be verifiable (hence see my second proposal). An argument could be made that unless the person has a way of verifying their action, there is no way to ensure that the desired outcome is achieved, so the instruction is invalid. (Separately, at least in your written example, it is possible for the person to be mistaken about the validity of their method!)

On the other hands, actions that are (machine) verifiable fall right within the scope of the Death Note (insofar as the verification instruments are also reliable, and verification happens within 23 days). I emphasize machine verifiable because the Death Note has various rules concerning other people, which I won't wade into. If "X starts his car and his car engine explodes" is valid, then so should "X keys a random string of inputs into a machine, machine returns True, then X dies."

So no we've limited the scope of DN-solvable problems to 23-day machine-verifiable problems. Now we just need to know what set of problems are 23-day machine-verifiable...

> it is possible for the person to be mistaken about the validity of their method

That's a good point.

> are 23-day machine-verifiable

Here's a followup question. Can a person be part of the machine?

For example, I might rephrase my earlier entry to be:

"X closes their eyes and types random keys, such that they form an algorithm for quickly factoring large primes. They email that algorithm to light.iskira@gmail.com and receive a confirmation that their algorithm is sound. At which point, they become so excited that they run into the street and are hit by a bus and killed."

My worry with that is that if there is some kind of force behind the death note, it might decide to just manipulate me instead of the person writing the algorithm. I don't want an evil genie scenario where I just accidentally write "yes, it works", instead of "no, it doesn't."

By extension, if we have a 23-day machine-verifiable problem, am I allowed to specify that the machine is not glitching or producing errors? In some cases, that might be more likely than finding the correct answer. Shinagami picking the most likely scenario that technically fits the criteria might be an effective way for them to reduce some abuse of the system.

This is the nerd version of asking the genie to grant you infinite wishes.

Similar to what the Mule can do in the Foundation series, except that he is able to push a person to have insights and make leaps in thought that the person would not otherwise have, at the cost of the person wasting away. He got the anti-nuclear field device that way and almost the location of the second foundation group.

> Worse, the deaths are non-random in other ways—they tend to occur at particular times! Graphed, daily patterns jump out.

I actually ran this kind of analysis on my use of Hacker News a couple of months back. It turns out that the patterns mentioned only really work if you have something that resembles a normal sleep schedule, which I clearly do not:

  12 AM  *********
  1 AM   ***********
  2 AM   *********
  3 AM   *******
  4 AM   *****
  5 AM   ****
  6 AM   ***
  7 AM   **
  8 AM   ****
  9 AM   *****
  10 AM  *********
  11 AM  ***********
  12 PM  ***********
  1 PM   ****************
  2 PM   *************
  3 PM   ***************
  4 PM   ************
  5 PM   **************
  6 PM   ***********
  7 PM   *******
  8 PM   **********
  9 PM   *********
  10 PM  ********
  11 PM  *******
(Each * corresponds to 25 comments; to my knowledge the vast majority of comments were made from PST)

As in the OP, it's pretty obvious when you are sleeping.

Sure, but would you believe that I'm now sleeping ~10 PM-7 AM? I'm not convinced there's enough information there to narrow down my location to anything smaller than a hemisphere.

I dunno, I think this still reveals some patterns:

Typically asleep around 5-8AM despite irregular sleep schedule, generally including 4 and 9 AM as well.

Frequently staying up to around 2 or 3 AM.

Lunch typically around 12, lull in work after lunch = more time to browse and make comments.

Dinner typically at 7. Spike in browsing HN after dinner while full, then focus on work or relax for a couple hours, before starting to drift off and get back to browsing HN around midnight.

Happy to hear if this is close or way off haha. But I suspect despite the patterns not being as clearcut, this is still enough to pinpoint your time zone within a couple hours.

> Typically asleep around 5-8AM despite irregular sleep schedule, generally including 4 and 9 AM as well.

At the time this data was taken, yes, I would generally be asleep at around 7-8 AM; usually that either meant I'd get up at 9 AM (if I had things to do) or past 12 PM (if I didn't).

> Frequently staying up to around 2 or 3 AM.

Yes, this is usually when I went to bed. Occasionally this would push towards 6 AM.

> Lunch typically around 12, lull in work after lunch = more time to browse and make comments.

I ate lunch around 1 or 2 PM, since I'd often wake up at about 12 PM.

> Dinner typically at 7. Spike in browsing HN after dinner while full, then focus on work or relax for a couple hours, before starting to drift off and get back to browsing HN around midnight.

Sounds about right.

> But I suspect despite the patterns not being as clearcut, this is still enough to pinpoint your time zone within a couple hours.

Could you do it if I didn't add the hour labels?

As a test, you could have removed the hours from your comment, shifted the lines by X, and asked HN to guess which hour the first line corresponds to

Yeah, I could have done that. If anyone else wants to play this, here's a script to print the timestamps of a particular user as Unix time: https://gist.github.com/saagarjha/6f6474fd212bd9904e6d7cc81f... the histogram above was generated by loading the output of that into a spreadsheet, skewing it by 7 hours to account for my timezone, and generating a ASCII histogram from that.

The ability to kill anyone one chose would be a far more terrifying weapon if used to instill fear. If someone used to this ability to kill individuals who didn't fit their agenda but kept anonymity about their goals, whatever changed they wished would only happen fairly slowly and haphazardly - those replacing the dead wouldn't have that much chance to produce different policies.

Also, even a bunch of seemingly random deaths with implications would likely produce terrifying fits of paranoia and retribution in the populace - remember humans can't not-see patterns in the thing they focus on, so a death-weapon would change the world this way - not sure if that what your psychopath wanted.

If an actor could kill "anyone nominated to head party X" then things would get ugly but even more, "do these things and you're immune" notes could certainly get people moving, especially if these were seemingly benevolent. "Random company heads die 'till CO2 production declines X amount" for example.

The Death Note could control what the victim did before their death. So Global Warming Kira could definitely have those random company heads die next to a piece of paper with "the deaths will continue until CO2 declines" written on it.

To clarify the Death Note rules discussion, this would be possible, but would need to be adjusted such that it's something the company heads would have reasonable done. So something more akin to

"Ceo Ceoson, overcome by immense guilt over environmental destruction, writes down 'Companies 1-16 are responsibl for the dying environment and we don't deserve to live.' having a heart attack."

It needs to double up as a stand-alone action as well as a threat.

Then you could write, "Before their death, the CEO of Gas Inc. will invent nuclear fusion, spread it to every market, and then live a long and happy life."

There are rules that would prevent that (they can only do things within their means, 1 month time limit).

What could work is an investment in such technologies, public announcements and then disappearing, maybe falling off a yacht or such a thing.

"Scientist X, working on Fusion, discovers method that advances research by decades, then dies of happiness and a caffeine overdose"

In the show this was limited to things that were both possible and that the target might realistically be able to do.

And really, it's what a death demon can do with some good old fashioned possession. So he can jump into a nearby truck driver and into the victim and have them meet in the middle of the road, but he doesn't necessarily know what the hell a prime number is.

There's a limit of a few days, if I remember correctly, on how long you can control people for before they die.

On "invisible death":

Regularity, order, and prompt obedience to command are qualities which, in modern armies, are of more importance towards determining the fate of battles than the dexterity and skill of the soldiers in the use of their arms. But the noise of firearms, the smoke, and the invisible death to which every man feels himself every moment exposed as soon as he comes within cannon-shot, and frequently a long time before the battle can be well said to be engaged, must render it very difficult to maintain any considerable degree of this regularity, order, and prompt obedience, even in the beginning of a modern battle. In an ancient battle there was no noise but what arose from the human voice; there was no smoke, there was no invisible cause of wounds or death. Every man, till some mortal weapon actually did approach him, saw clearly that no such weapon was near him. In these circumstances, and among troops who had some confidence in their own skill and dexterity in the use of their arms, it must have been a good deal less difficult to preserve some degree regularity and order, not only in the beginning, but through the whole progress of an ancient battle, and till one of the two armies was fairly defeated.

-- Adam Smith, Wealth of Nations


There are three uses of the word "invisible" in the text. Two refer to invisiblbe causes of injury or death, and their psychological impact.

This is exactly what happens in totalitarian regimes. I know the CCP does this in China, but I assume Hitler, Stalin, Mao, maybe the Saudis, did/do this.

To a lesser extreme, so does Singapore.

Not that I have a reference anymore, but there used to be articles in the state controlled newspaper mentioning the fuzzy bounds of allowed political discourse akin to I'll know it when I see it and if you so happen to cross this boundary, well that's your own fault for getting involved in politics at all, you should've stuck to simple peasantry.

Characters in fiction regularly behave sub-optimally even if they're set up as geniuses, for the most mundane meta reason: It's a work of entertainment, and the plot wouldn't get interesting otherwise! We should always be very aware that any fiction, even ones that attempt speculative simulations, is always shackled by the need to keep being entertaining and interesting -- there are craptons of concepts and possibilities that _can_ occur or _will_ become deeply relevant to humanity but hardly ever gets explored in fiction because they're boring to read or write about.

I really enjoyed Harry Potter and the Methods of rationality. I understand not everyone would like such an intelligent main character though.

One thing the author seemingly didn't address - the overlap in entropy between the different mistakes. For example, Mistake 2 in timing leaked 6 bits, but those same 6 bits got leaked as a part of Mistake 4 in using confidential police information. Mistake 4 on its own is possibly the worst, but if you consider what L already knows, it's only 5 incremental bits lost.

Despite that, loved the article.

The two mistakes corroborated each other and the second was only possible to observe based on the refining characteristic intrinsic to the earlier mistake. Additionally the second wouldn't have necessarily invalidated the first six bits, though it would strengthen refinement along a different path.

Information is defined as relative to existing sets. Mistakes can change how many bits they leak if other mistakes didn't happen - but they did. And in a broader sense, most of these mistakes are dependent on a prior mistake: the FBI team wouldn't've been there if not for the Tailor mistake, and so on.

This was well cited with a lot of fun media references. The author makes a few interesting points on rationale for randomization, calling known unknowns and confounding factors as something one could plan for and avoid.

Anybody know of other writers that do research and writing in this style? I quite enjoyed it.

Gwern Branwen has a large body of work on a broad range of topics. You could go to the top of his web page https://www.gwern.net/ , pick a topic that interests you, and start reading.

I admire his site, but for some reason, I found his content very hard to digest. I can read if someone links to any particular article, but browsing through the index is very overwhelming to me.

I don't know if it's just the sheer amount of content that feels intimidating or anything else (like reading long paragraphs of sans-serif fonts on screen).

"Anybody know of other writers that do research and writing in this style? I quite enjoyed it."

This is what immediately comes to mind:


This website really impressed me - a lot of interesting stuff to read, for sure.

Agree wholeheartedly. It's rare to find a website today that is so abjectly different than the rest of the web.

Abject =~ despicable

I find 'abject' an amusing adjective to apply. It makes one imagine the author writhing on the ground in despair after loading a standard website, and despondent - o tempora! o mores! - resolving to be different.

(Not entirely wrong, as it happens.)

It's a very cool format, for sure.

Scrolling through it quickly makes me feel like like I loaded a webpage that hasn't loaded images properly and show a little X icon instead.

There aren't any images, so I'm not sure what you mean. Are you referring to the sidenotes?

As an admittedly not very smart person, the sidenotes do create a very claustrophobic effect on me. Maybe it's different when one is reading the main article and glancing at the notes, but when randomly scrolling or pgdn-ing to skim the article, it's overwhelming.

If it's true that privacy is dead, how come Satoshi is still unknown?

Satoshi hasn't posted in almost 9 years. If he was still active, it's much less likely he would still be unknown.

Maybe he was discovered and killed?

The deathnote game is played iteratively. Satoshi ceased publishing before running out of bits.

This is true. I would very much like an article in this style examining Satoshi and how many bits he theoretically had left.

I have written one but haven't posted it for obvious reasons. In Satoshi's case, it's vexed by the problem that it depends heavily on what datasets you have access to. For example, the California IP leak reduces Satoshi's anonymity to perhaps 2 or 3 bits... but only if you know the location/identity of everyone in California and also where that IP was assigned to at that time.

Baseless gut-guess: Any move clearly associated to Satoshi, including use of any of the Bitcoin value, would receive such unrelenting scrutiny that there are only two likely outcomes.

A) Satoshi would be identified as a unique individual and any number of bad things would happen since they are such a high value target.

B) Satoshi would be identified as (effectively?) a pseudonym for one or more nation states and whatever value they derived from creating Bitcoin would evaporate.

Given we are currently in a superposition between states A and B, as well as a possible state C where the pseudonym Satoshi is either effectively dead (lost the credentials) or actually dead (end of life); one can infer that the value of Bitcoin continuing to exist as it is either outweighs any change of state or that such a change of state is no longer possible.

Stylometry only works too an extent with helping to identify what's what. I could just stylometry to show that this document was written by a professional screenwriter, but it writing style isn't unique enough to identify one person out of billions. Gwern has an excellent post about this.

I remember a really impressive demo where someone found all the alt accounts of a HN user using stylometry: https://news.ycombinator.com/item?id=17944484 The account didn't have that many comments on it (maybe ~1000 karma)

He also claims to have used the same algorithm for identifying company insiders for trades (see parent comment in that thread for context).

I don't think he was bullshitting even though his tool isn't public because he has a different NLP project that works reasonably well (https://hnprofile.com/ which can find users based on their interests).

Agree 100%. I'd believe that the DHS has identified Satoshi, but very very skeptical that they could use mass email collection to do it. Stylometry could work though if you limited your search to the set of people active on niche pre-Bitcoin lists/forums where Satoshi was likely to be active under a different name.

It depends on how much data you have. When you have everyone's emails you can rank everyone by closest match and narrowing parameters.

You and I might not know who they (1..n) are, but maybe some sleuths have figured it out, and not leaked the fact that they have.

Is he? He leaked a lot of writing samples. Unless he took great care to anonymize his writing style (and I'm not even convinced that it's possible to completely anonymize a writing style without scrambling the ideas themselves), it's possible (perhaps likely) he leaked enough bits for a state actor to track him (or her) down.

I'd expect they'll be AIs in 2040 that'll be able to figure out who he is by writing samples, at least. And the rest of us here on Hacker News, too.

This is up there for one of the most thorough proofs that God (if she exists, for most meanings of exists) cannot interact with the universe in ways that contravene statistical mechanics.

Someone on HN linked to a list of the best anime - and I can't find it now, because I foolishly thought I would remember the name of the website.

Anyone know what I am talking about? The "winning entry" was, I think some kind of pretty abstract anime story. If Death Note was on the list at all (the list may have been compiled before Death Note was aired), it was not on the top of the list. The list was written by a single author and it had extensive long-form style commentary.

I think you're a few bits short of specifying a website here

You could try https://hn.algolia.com/

Death Note is great...


Death Note is based on the "keep your enemies close" that turns up a lot in Japanese Culture. For instance, more than once in Sailor Moon the bad guys wind up embedded with the good guys, and the same thing happens in Tales of Symphonia. Thus it is almost necessary for the plot for Light to move in with L.

Light's mistakes are part of the plot, not an error in it. He eventually loses as the Shinigami predicts that he will be the one writing his name on his Death Note. L also sacrifices himself knowing that he should close down on Kira and Near eventually puts an end to it. Yet my issue was with Light killing using the bag of chips with the TV inside, what kinda TV could fit there and render visible information with low light in a dark surveilled room?

Ah but if he hadn't done that we wouldn't have gotten the classic "I'll take a chip.... AND I'LL EAT IT!" line which really couldn't be more critical to Death Note.

On a more serious note I don't think this is trying to say the show is wrong for having these mistakes just that the character could have done better and thinking about how is fun and allows you to examine just how much information leaks from seemingly benign sources.

This might be the nerdiest article I've seen.

Also a great illustration why CoinJoin is not a complete transaction privacy solution for Bitcoin.

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