Hacker Cinema
Tora! Tora! Tora!
1970 · Film · 144 min
Directed by Toshio Masuda, Kinji Fukasaku, Richard Fleischer
In the summer of 1941, the United States and Japan seem on the brink of war after constant embargos and failed diplomacy come to no end. "Tora! Tora! Tora!", named after the code words used by the lead Japanese pilot to indicate they had surprised the Americans, covers the days leading up to the attack on Pearl Harbor, which plunged America into the Second World War.
3 positive5 neutralfrom 8 commenters
Where to watch
No streaming data for this title.
8 comments on Hacker News
I dunno, 2015, I'm a little weak on this year: It Follows, Creed, Ex Machina, Max Max: Fury Road, Sicario, The Hateful 8. I'd watch those again with someone any time, and recommend all of them often. Several on my to-watch list most of which I expect to be good, like Queen of the Desert, While We're Young, Slow West, and The Overnight, just haven't made it to them yet. Not an exhaustive list, just gleaned from some titles I have at hand.
2016: Green Room, The Nice Guys, Hail Caesar!, The Neon Demon, Swiss Army Man, Hunt for the Wilder People, The VVitch, Train to Busan, Shin Godzilla, Moana (hey, I like this one), Arrival. Moonlight's on my to-watch and is supposed to be really good.
2017 (I've done OK on catching up with these!): The Lost City of Z, Dunkirk, Low Life, Good Time, Logan Lucky (absolutely slept on, kills it as a feel-good lightweight small-stakes heist movie), Blade Runner 2049, The Death of Stalin (I liked this less than a lot of folks, but given how widely-loved it was, I'm probably the idiot here), One Cut of the Dead, You Were Never Really Here. The Planet of the Apes movie from that year, plus Phantom Thread, and The Little Hours are to-watch for me and come highly recommended.
2018: Annihilation, Isle of Dogs, Upgrade, Sorry to Bother You, High Life, Eighth Grade, the Suspiria remake. To-watch that I expect to be good include Champion (Korean arm wrestling movie—there's another movie by the same name that year), First Reformed, BlacKkKLansman, The Favourite, The Wolf House, Climax, and some others.
2019: HUGE year for the particular (small) set I'm seeing on my list, including a ton to-watch but a bunch I've seen. Uncut Gems, The Lighthouse, Knives Out, Little Women, JoJo Rabbit, Ready or Not, Parasite, Midsommar, The Art of Self Defense, maybe Marriage Story (but if you've seen one Baumbach movie, you've kinda seen them all, and I'm not sure I'd put it above The Squid and the Whale). Midway's a well-above-average war movie and pairs great with Tora, Tora, Tora! as a crazy-long double feature in a really fun way. To-watch list is nuts and I really need to dedicate a month or so to filling out my watched-list for this year: First Cow, The Irishman, Portrait of a Lady on Fire, Bacurau, Once Upon a Time in Hollywood, The Farewell, The Souvenir, Blow the Man Down, The Vast of Night, Her Smell, Funan.
It basically just keeps going like that, year after year, and I've barely even tried to dredge up good movies for most of those years, the bulk of it's just stuff that's risen to my attention one way or another, and I'm terrible at keeping up with foreign film especially. I also left off some that a lot of folks would probably include, like at least one Mission Impossible (aside from MI2, I think these are all pretty good action movies, though quality varies a bit) and Avengers: Endgame which, opinion on the rest of Marvel aside (I think it's mostly kinda lazy crap? But was basically entertained for most of them regardless, so I guess I can't complain too much) was a hell of an event. Also Black Panther, which everyone loved but I was pretty meh on (I hate the entire end fight, and it's looooong)
Veeery interesting. After realizing that Tora, Tora, Tora! and Midway might make a good double-feature (I tried it, and, they do!) it occurred to me that it might be possible to assemble a film-based curriculum to teach a great deal of the history of roughly 1933-1948, covering the lead-up to and immediate aftermath of the war, in a way that's entertaining while being more informative than misleading. There are thousands of films covering the time period from dozens of countries, and lots of those stick reasonably close to historical events, so it might work out.
The hard part, I think, would be tracking down films that give a good sense of the causes and course of more-obscure things like Italy's invasion of Ethiopia. You'd need to find two or three good films on that. Spanish civil war? The invasion and occupation of Poland? The political maneuvering between the Nazis and Soviets before they went to war with one another? The Winter War? These have to be covered by a few films that could act to "teach" the events, but I don't know what those films are and bet most are non-English and not well-known in English, making them harder (for me) to track down.
Alien vs The Andromeda Strain vs Taking of Pelham One Two Three vs The Godfather vs American Graffiti vs Animal House vs Apocalypse Now vs Tora! Tora! Tora!
Honestly I don't see much connecting "70s-ness" thread between these. You can kind of take a guess at the era by the color tones, or for those movies with a contemporary setting, the style of clothing and cars. But otherwise there is huge range in old movies.
I've got this theory that although technology advances and social mores change with different social standards being imposed on art (e.g. Hayes Code), artistic merit is more or less evenly distributed through time. That is to say, old things are about as likely to be meritorious as new things, there's no general trend of newer = better. However the general public realizing this would be bad for business, new productions don't want to compete with the best of old generations, and so there are efforts to persuade people that old content is unwatchable and a waste of time to even consider. People are strongly encouraged to favor new content over old, for instance by having an averse "eww icky" reaction to black and white movies or even technicolor. Technical developments in film stock, SFX, etc are touted as being extremely meritorious and are heavily emphasized in the promotion of new movies because this is the competitive edge for modern productions.
> I feel there are plenty of good shows on Netflix.
Not sure who else feels this but for me:
There's a lack of "regular tv shows". I knew it was over when they got rid of The Office, that was pretty much the only reason I kept Netflix.
HBO Max, Disney Plus, Prime Video, AMC (temporary) all have the shows we regularly go to. Sopranos/Frasier, Simpsons/Star Wars/National Geography, The Office/Friends, Breaking Bad/Better call Saul (just need to see how it all ends).
Heck even Prime managed to bag many of the shows that used to be on Netflix. Aliens, Tora Tora Tora and even Independence Day is on Disney Plus. Lot of movies outside Netflix.
Netflix just has these one off good shows that frankly doesn't attract across the demographic. Sure there's something for everyone but its lacking those "must have and easy to access binge reruns".
The Opana Radar Site on the north corner of the island detected the Japanese aircraft approaching, and the radar operators there reported it to their command. However this report was confused for a formation of B-17s that was scheduled to arrive from that general direction. The size of the detected formation, which might have revealed the confusion, was not properly communicated.
This incident was a subplot in the 1970 film Tora! Tora! Tora!, which may be where you heard it.
Part I
Here at HN I'm trying to be anonymous but on anomaly detection make a contribution.
The basic idea of my research is for positive integer k, use k-nearest neighbors or, really, any of a wide variety of such measures, metrics, functions, windows however want to regard them.
So, right, are on the way to some approaches to multi-dimensional probability density estimation.
The main contribution of my research is the applied probability derivations to show how to calculate, adjust, set, etc. the false alarm rate. In practice, false alarm rate unknown, out'a control, often too high is a biggie problem -- e.g., can cause staff just to ignore too many alarms. E.g., see the movie Tora, Tora, Tora about the 12/7/1941 attack on Pearl Harbor where the US actually had just what the heck it needed for early detection, early enough to get fighter airplanes in the air, maybe early enough to get the ships out of the harbor, etc. -- radar on top of a hill. And the radar operators DID get the detection and DID report it, but the chain of command assumed that the detection was a false alarm. Big, huge, gigantic bummer. The US came within the thickness of an onion skin of getting the US airplanes and ships out of harm's way, shooting down much of the attacking airplanes, and having the US carriers, already at sea, attacking the attacking carriers when all their airplanes were off over Hawaii. The US could have sustained little damage and sunk all six of the attacking carriers.
Having a good anomaly detector and ignoring it can be a bummer.
But if know what the false alarm rate is and know that it's low, then maybe the chain of command will take a detection more seriously and, then, take the next step -- diagnosis.
Sometimes the spook business, say, Soviets in England, would watch the comings, goings, lights on late, etc. at British military HQs and try to detect anomalies that would indicate war. Sure, for such detection need both distribution-free and multi-dimensional with good control over false alarm rate and good evidence of a relatively high detection rate. So, sure, the US CIA should want my work. I wrote them -- never heard back.
Once I did give a talk on my work at the Institute for Defense Analysis (IDA), think tank resource for the Joint Chiefs. Some in the audience mentioned credit card fraud detection. Okay. But there should also be national security applications.
Once gave a talk on the work at the Johns Hopkins University Applied Physics Lab (JHU/APL) -- I might as well have talked about a recipe for pancakes.
Once I gave a talk on the work at the NASDAQ site in Trumbull, CT. I got a nice tour, a nice lunch, and a nice tote bag.
I gave some talks on this work to my management chain at IBM's Watson lab -- the suits were afraid I might have some good work they didn't understand, couldn't control, and couldn't throttle. They were PISSED.
I presented my work to some IBM people outside of Watson Research -- they liked the work, but Watson Research was PISSED.
For the applied probability stuff in my research, for a detection of an anomaly can also report the lowest false alarm rate at which the data would still be a detection. Then right along can graph these lowest rates and use them as a heuristic about detection seriousness. Net, getting a handle, knob to turn, knowledge of, etc. false alarm rate is important.
The applied probability stuff is based on some cute symmetry in the calculations. The symmetry, then, gives some measure preserving transformations (transformations that don't change the probability distribution). So, end up with a finite algebraic group of such transformations. Then sum over the group and, presto, bingo, get what need to know false alarm rate.
To adjust the false alarm rate, adjust the diameter, cut-off, whatever, of the k nearest neighbors or other metric, convolution window, etc. Of course all of this is conditional probabilities conditioned on the historical data used where assume that the system being monitored when the historical data was collected was healthy. How to confirm healthy? Let the data age and see if there are any other, later symptoms of problems from other means, e.g., angry customer complaints, malware, attacks and intrusions, etc.
So, have a case of, essentially, being brief here, the conditional expectation E[Y|X] where Y is the random variable of detections and X is the historical data. Then of course E[E[Y|X]] = E[Y] which says that have set false alarm rate "exactly" in the long-run or, if you will, conditionally exactly in the short run.
A result is that the naive way to evaluate false alarm rate from the historical data does, for lots of data, converge to the right answer. Proving that, however, was a bit tricky.
Achieving the Neyman-Pearson best result needs lots of data from both healthy and sick systems, and in practice the second is in short supply.
The Neyman-Pearson result is intuitive, like investing in real estate: Buy plots of real estate that give the best return per dollar invested until have invested all the available money.
Well, suppose the data observed is, for the set of real numbers R and some positive integer n, in R^n. Then partition R^n into little patches. Then in each little patch, assume that data in that patch would raise an alarm, and, then, from the data (need lots of data, especially as n grows) get the ratio of the detection rate over the false alarm rate. Then false alarm rate is really like the money you have to invest, and detection rate is your return on that investment. So, sure, you buy the patches with the highest ratio until you have spent, invested, all your false alarm rate, money.
Can find proofs in terms of freshman calculus. I wrote out a quite general proof in terms of the Hahn decomposition from the Radon-Nikodym theorem in measure theory.
So, right, if have only finitely many patches in R^n, then have essentially a knapsack problem, IIRC, in NP-complete (although likely not a concern in practice).
Can write out a little derivation, mostly just Fubini's theorem (the measure theory version of the calculus interchange order of integration) to show in a cute sense of the distribution of sick systems being translation invariant (right, a bit much to assume -- but here I'm typing quickly from memory of what I did 20+ years ago) then my techniques give, for the assumed false alarm rate, the highest detection rate possible. So, this is a simple, quick and dirty, data-free, crude approach to the full Neyman-Pearson result of best detection.
The "creepiness factor" of targeted advertising.
I am completely cool with advertising, as well as Real Life targeted advertising such as direct mail.
However I am profoundly offended by targeted online advertising. For example a few days ago I was scouting around for a book on purely conceptual mobile website design. I like APress' books on conceptual iOS App design, so I wandered around APress' site, eventually to come up empty-handed.
Now, when I log into facebook, absolutely the whole fucking time is an ad for APress books.
Look I think APress is a right chap but I don't want him following me home from the WiFi spot late at night!
However I realize now that I have a realistic method to make A Compleat Clusterfuque of Facebook's ad tracking:
I have so far avoided entering any of my favorite books or movies. I also - from time to time - advise my FB friends to avoid it themselves. Few of them were previously aware such tracking was possible.
Here Are A Few Of My Favorite Things:
Movies:
The President's Analyst 2001: A Space Oddyssey Logan's Run (all those comedy flicks that star Leslie Nielsen) Bound For Glory Dr. Strangelove, Or How I Learned To Stop Worrying And Love The Bomb Tora! Tora! Tora! Most but not all the James Bond flicks Very few of the sequels to anything
Books
Zen and the Art of Motorcycle Maintenance All Quiet On The Western Front The Hobbit (but not Lord of the Rings) 2001: A Space Oddyssey (quite different from the movie) (most but not all of Kurt Vonnegut's novels) Amateur Telescope Making Foundation Trilogy Stranger In A Strange Land Time Enough For Love
I need to catch some ZZZs soon so I'll cut this short.
More or less what I'd do is list a bunch of favorites, as being a whole bunch of completely unrelated books and movies, absolutely none of which I'd actually seen nor read, generally having to do with topics that either do not interest me or even offend me.
Daniel Steel is quite popular. I read a few pages of one of her novels it was well-written but I could not conceive of actually reading the whole book.
While this wouldn't stop FB from showing me ads, it would result in a lot of wasted time and effort showing them to me.
What really gets me down, though, is that the vast majority of ads that I see, are offering me attractive women of my own age.
I am fifty-one years old. The one small comfort I take form my divorce, is that Bonita did not want to have children.
As yet, Facebook has yet to offer me a woman who might bear my child. Figure That One Out: many if not most of my FB friends have kids, they are always posting their photos, singing their praises &c.
I always "Like" their posts, commonly reply, sometimes share yet FB has yet to clue in to flogging a way for me to continue the non-quite Eternal swim in the gene pool.
:-(
I was working as on a corporate dev team that hit a goldmine at one point, causing management to freak out and go on a hiring spree. They brought at least three wolves on board within a period of a couple months. One of them reminded me of the portrayal of Kameto Kuroshima in the 1970 film "Tora Tora Tora". He would think and think for hours on end, then he'd emerge from his cubicle (all sweaty and dramatic) with a Plan. At that point everything had to change because everyone knew his experience levels blew us all away. When he attended meetings, it was like he was five steps ahead of everybody else, and later I realized that it was because the meetings were usually called as a result of something he decided should change. I could easily see most of the people above him being fired if this guy ever got pissed.
One of the other wolves was an experienced college professor who was brought in as a team lead, but who for all intents and purposes was an SVP in the organizational hierarchy. He was financially independent at a fairly young age and basically gave inspirational speeches and CS lectures instead of meetings. We were all freshmen to this guy, and he was light years ahead of our experience levels.
I liked these wolves, if that's what they were--they gave me hope that somewhere an executive was really freaked out about making the right call at a critical time.