Twitter Based "Ted" System Warns of Earthquakes Earlier

hypnosec writes "A Twitter-based system managed to detect the earthquake off the Philippines before any other advanced spotting systems being used by Seismologists. The U.S. Geological Survey uses the micro-blogging site to quickly gather information about earthquakes around the globe through the use of a system — Twitter Earthquake Detection (TED) — which beat out USGS's own sensors on Friday when it came to detecting a 7.6 magnitude earthquake off the Philippine coast. The TED system gathers earthquake related messages (Tweets) in real time from Twitter. The system takes into consideration various parameters like place, time, keywords, and photographs of affected places where tremors have been detected. Online information posted by people — Tweets, in this case — can be picked up faster by researchers, compared to scientific alerts that may take up to 20 minutes."

Comments

Looks like another XKCD that works

[spaceplanesfan]

http://xkcd.com/723/

Re:Looks like another XKCD that works

[ColdWetDog]

I'd be worried more about the timing...

On August 31st, Twitter became self aware and launched an earthquake in the Philippines to incite a counterattack against the humans, who in panic, tried to log on to Facebook.

Be afraid.

Be very afraid.

Re:Looks like another XKCD that works

[sortius_nod]

Uhhh, that's what the "TED" system is doing. Seriously, read the fucking summary at least.

Re:Looks like another XKCD that works

[JustOK]

Seriously, read the fucking summary at least.

Why, what did it say?

Re:Looks like another XKCD that works

[GoodNewsJimDotCom]

Whoa, chill out rage man. I read it and the "TED" feed. I'm suggesting upgrades. TED only says the locations of where the earthquakes are. I was thinking of more precise engine which gives exact long/lat of it, and even takes into consideration your smart phones long/lat. Sure TED looks nice in that it can tell you ahead of time, for most of the time, it's gonna be tweeting about EarthQuakes who aren't around you, which is like crying wolf and your alarm goes down. I was thinking of a smart phone ap that read from TED for your own specific encounter, as to not give out lots of false alarms etc etc etc. Does this make sense to anyone? It could be called "Earthquake Ap", and only alarms you if an EarthQuake is withing X hundred miles of you.

Re:Looks like another XKCD that works

[AK Marc]

Uh, no, it doesn't. I've felt an earthquake 200 miles from the epicenter when others 100 miles from the epicenter felt nothing. The plates were such that it was felt to the south, not the north. TED doesn't try to predict *your* experience, but give you info and let you figure it out for yourself. Warning based on proximity, not based on direction and strength.

the downside

[Anonymous Coward]

Is that if you get 50k Anonymous, you'll end up with enough false positives to make the system useless.

On the other hand, having idiots tweet prior to taking shelter can only improve the species.

Re:the downside

[bennomatic]

I was thinking a twitter flashmob in this context could have interesting repercussions..

Confusing data and information

[vlm]

Confusing data and information.

The number of tweets with the word "earthquake" in them is a raw piece of data.

A USGS trained analytical geologists opinion of magnitude / depth / time is much higher level information and is going to take a couple minutes at least to think about it. I suppose if the on duty guy is in the can when a earthquake starts, or even worse if you have two guys who start arguing before one finally calls the boss to resolve the fight....

The standard /. automotive analogy is that a bunch of tweeter twits will always be the first to "report" a car crash. It'll take a couple hours at least for the accident investigation team to gather all the legal evidence, and at least days until a judge / jury convicts. I'm sure the twits are always going to be faster than Judge Dredd or whoever.

Re:Confusing data and information

[Hazel Bergeron]

"Training" is just a class-obsessed synonym for "education".

And the lack of hyphen between "USGS" and "trained" should help resolve the typically awful Slashdot proofreading.

Re:Confusing data and information

[catmistake]

A USGS trained analytical geologists opinion...

yeah... but buddy, geology isn't a trade, per se... geologists aren't so much "trained" as they are, uh... educated... and generally, it isn't the USGS that is educating them so much as they are, uh... employing them.

You have no idea what you're talking about... geologist are born. A friend of my cousin's was drafted by the USGS, trained at Top Rock, and did 2 tours in the Grand Canyon, and saw a lot of interesting sediment, so I know what I'm talking about!

On March 3, 1879 the United States Geological Survey established an elite school for the top one percent of its geologists. Its purpose was to teach the lost art of earth's physical structure and substance, its history, and the processes that act on it and to insure that the handful of men who graduated were the best geologists in the world. They succeeded. Today, the USGS calls it the Department of Geology (found at various universities). The geologists call it: TOP ROCK.

Iceman: You two really are rockboys.

Maverock: What's your problem, Cavenski?

Iceman: You're everyone's problem. That's because every time you go down in the cave, you're unsafe. I don't like you because you're dangerous.

Maverock: That's right! Ice... man. I am dangerous.

Re:Confusing data and information

[VTI9600]

No, I'm sorry, but your analogy is flawed. What actually happened was that Judge Dredd used his Lawgiver-II grenade launcher to destroy a car that was ilegally parked. Within seconds he triumphantly announced that justice had been served via his megaphone. Minutes later, other cars in the megacity were trembling with fear. It wasn't until several hours later that the coucil of judges came up with an adequate rationaliztion for what, on the surface, appeared to be an egregious misuse of police force.

FTFY.

Re:Confusing data and information

[Hazel Bergeron]

I ain't goin' in no cave.

Re:Confusing data and information

[Hazel Bergeron]

Thank you, Cap'n Obvious.

Re:Confusing data and information

[urusan]

I'd say that education and training really are two very different things. Training is specific to an individual task or set of tasks while education is broader and has more to do with cultivating independent thought.

In this example, the training is specific to earthquake detection. In theory, you could train someone to just detect earthquakes given a set of input data (from seismometers and such) and leave them worthless at anything else geology-related. In practice though the specific training rests on top of a deeper education (at least for most scientific and analytical jobs). As another example, in programming when you learn a new technology or language you're training in those things, despite the fact that this training rests upon a deeper education.

It should also be noted that a lot of what we commonly consider "education" is actually training in disguise. For instance, arithmetic and calculus are just methods to produce a desired result. When we spend time learning how to do these methods by hand, we're doing training. It's no different from training to do a sport, as the end goal is to be able to fluently perform a desired action so you can perform on command. Even highly educated individuals spend more time training than being educated.

Education is more meta and has more to do with the effective use of training (including training one does not personally have) as well as some even more meta concerns (which is to say less practical stuff that enriches you as a human being, but is really only interesting internally). From a practical standpoint, an educated person can effectively organize their own training to great effect as well as (hopefully) organizing the training of others into a cohesive whole (though leadership has its own required training).

The "class-obsession" you talk about has more to do with how training and education are used in practice. Most non-upper class students receive the minimum required education, with most of their "education" being composed of training. In egalitarian societies, a real education is additionally offered to the brightest non-upper class students, as they are very useful when they are properly educated. I should know, I was born into a working class family and received both classes of education at different times.

A bit off topic, this is why I'm worried about the economic and social impact that weak AI will have. We're able to train computers now and only those with a real education will have value to add as weak AI becomes smarter and more common. Since a large percentage of the populace is only marginally educated, this does not bode well.

Re:Confusing data and information

[RockMFR]

These two quotes from the article are confusing to me: "We do have sensors and it usually takes about five minutes before the sensors will see the earthquake" and "scientific alerts can take between two and 20 minutes".

Why would a seismometer take so long? Can someone explain this?

As far as I know, the USGS website has data in realtime and doesn't necessarily wait for any kind of human verification. For example, within approximately 15 seconds of the Virginia quake last year, I checked the USGS site and they already had it in their list of recent quakes. For other quakes I've seen reported on Twitter, the USGS site has always had data up at least as soon as the first tweets I've seen.

Re:Confusing data and information

[AK Marc]

Since they dropped the "how much did the earth move"measurement to "how bad did it feel" measurements, twitter is more accurate than their seismographs.

Re:Confusing data and information

[AK Marc]

I've had different results. I logged in to the university's site, and their readings showed an earthquake, but didn't call it one for quite some time. The shaking data was instantly available, but "earthquake" reporting takes a trained scientist to look at the obvious squiggles and say "earthquake".

Re:Confusing data and information

[Neil Boekend]

All humans are born (bar measurement errors)

Re:Confusing data and information

[Hazel Bergeron]

Thanks for your response.

Training is specific to an individual task or set of tasks while education is broader and has more to do with cultivating independent thought.

So training is bad education? Every talent can be reduced to (possibly multiple simultaneous) walks of a graph, wisely choosing from a set of alternatives at each node.

In theory, you could train someone to just detect earthquakes given a set of input data (from seismometers and such) and leave them worthless at anything else geology-related.

True. But it's an arbitrary human decision to categorise various things under "geology", so it's equally arbitrary to suggest that having a general knowledge of "geology" is education while only knowing one subset is "training". The training might in fact be regarded as education stimulating statistical thought.

It's no different from training to do a sport

On the contrary, some (not all!) sports played at high level require a sharp mind: to analyse the field, to predict, to respond at the right moment, etc. It is the cunning, unpredictable players who will shine. Sportspeople also have to be physically fit and nerds remember the jocks at school who bullied them, so may assume that sports aren't really for the mentally fit.

Education is more meta and has more to do with the effective use of training [...]. From a practical standpoint, an educated person can effectively organize their own training to great effect as well as (hopefully) organizing the training of others into a cohesive whole

OK, but good training encourages self-improvement. Back to the tree-walk, the organisation of learning ("learning how to learn") is itself a talent which can be learned. Looking at education this way, is it not just another subject to be trained? Components of this training are part of all good "training"/"education" in any field.

The "class-obsession" you talk about has more to do with how training and education are used in practice.

It's true that the rich have better access to education/training, as they have better access to anything. I maintain that "training" is used in order to denigrate the education of certain classes. I can understand the distinction between, say, a more "vocational" education and a more "academic" education, where the former involves a different skillset. But I think it's a mistake to suggest that they're not both education. Indeed, what's happened in the UK - with everything turning into a University and everything becoming a degree - is the result of one group attempting to tilt at this windmill, and a cynical Tory party responding to it in a self-serving manner. "Let's educate everyone!" pretended Thatcher. But most people were already being educated. Now, in an attempt to make one technique fit all, far fewer are receiving an education.

In egalitarian societies, a real education is additionally offered to the brightest non-upper class

In an egalitarian society, there is certainly no "upper class" from which to differentiate a "non-upper class", and education is not only offered to the bright.

as weak AI becomes smarter and more common

This is only a transitional step to creating AI more intelligent than humans, surely? If AI is a threat then it is a threat to everyone. You may say that at some point AI becomes strong enough to deserve rights, but humans are great at denying rights to things which look sufficiently different.

Since a large percentage of the populace is only marginally educated,

As someone who went to an uppity private school on scholarship, I'd have to agree: the people around me were only marginally educated. Boarding school was training, if you like, in saying the right things to the right people.

this does not bode well.

In the long term, our economic system does not bode well. Although his thoughts about alternatives were somewhat(!) unscientific, Marx's predictions about capitalism were sound and have mostly been realised.

Re:Confusing data and information

[Impy the Impiuos Imp]

It would be interesting if an analysis of tweets just before the earthquake indicate anything odd in the subconscious of the twitterers.

In any case, if you did an analysis of, say, discussions of kids with autism, you would see a hotspot, literally (and this is known to skeptics) around Hollywood, and think there something dangerously wrong there.

Re:Confusing data and information

[vlm]

I subscribe to one of the RSS feeds at http://www.gdacs.org/ and sometimes it takes them awhile to push out earthquake alerts. The raw data has already been chewed on a bit and for earthquakes they provide depth, magnitude, a map of where the fun is, some rough guess of affected population...

There are other similar services, this is just the one I RSS subscribe to.

Re:Confusing data and information

[vlm]

anything odd in the subconscious of the twitterers

You mean unusual not odd. twits are already pretty odd people all the time.

Re:Confusing data and information

[urusan]

The main issue here is that we're using different definitions of education. I'm more specifically thinking of a liberal education. http://en.wikipedia.org/wiki/Liberal_education In particular, when I say that someone has an education I'm really talking about the state that a liberal education strives to create.

So training is bad education? Every talent can be reduced to (possibly multiple simultaneous) walks of a graph, wisely choosing from a set of alternatives at each node.

No, it's a compliment to education. It's what makes one fluent at whatever they're doing. For instance, when I typed that last sentence, my training in the English language allowed me to realize that outcome quickly and efficiently. Without my training, I'd be constantly stopping to look up words and remember grammatical rules. It would be slow and tedious. Training is critical for any function that isn't purely instinctual. Most of what you learn in any realistic educational system is training.

However, learning to think independently is much harder. If one cannot think independently, one can't learn to think independently by themselves. Worse yet, it can't be externally trained. How can one learn to be an independent person when there is a teacher (or book) to be dependent upon? The best a teacher can do is set up an environment that is likely to lead to independent thought. Even in the best situation there is no guarantee it will work. In fact, this is true even assuming we're deterministic machines like computers, because questions about independent thought gets into self-referential territory and proving stuff about it becomes undecidable (or intractable if you want to get really technical, as we don't have infinite memory...but the size of your tree for a human is exponential with an input number somewhere around the number of neurons in the human brain...easily more than 2^100,000,000,000 which is way more than the 10^80-ish atoms in the universe).

a general knowledge of "geology" is education

A general knowledge of geology is (extensive) training.

By the way, I have known several scientists and programmers with a relatively low capability for independent thought. It's much rarer than in the general populace, but it's still there. Also, I've noticed that certain national educational systems create large numbers of graduates that can't think independently. I was recently in a training program with nearly 40 programmers from such an educational system and it was almost comical how few of them could do anything substantial without external support.

On the contrary, some (not all!) sports played at high level require a sharp mind: to analyse the field, to predict, to respond at the right moment, etc. It is the cunning, unpredictable players who will shine.

This is true, but I used it as an example because: 1. most people call sports preparation training and 2. it's undeniable that it takes a LOT of training to succeed at a sport.

OK, but good training encourages self-improvement. Back to the tree-walk, the organisation of learning ("learning how to learn") is itself a talent which can be learned. Looking at education this way, is it not just another subject to be trained? Components of this training are part of all good "training"/"education" in any field.

Learning to learn is another one of those tricky self-referential skills (how do you learn how to learn in the first place when you don't know how to learn?), so it can't be deterministically trained. The best we can do is set up the environment so it happens most of the time.

In an egalitarian society, there is certainly no "upper class" from which to differentiate a "non-upper class", and education is not only offered to the bright.

I guess that's true in the strictest sense, but if so there is no such thing as an egalit

Guy tweets friend having heart attack...

[Hazel Bergeron]

...praised for tweeting near death experience!

(Medics arrive 5 minutes later, revive friend.)

News at 11.

Re:prone to

[History's Coming To]

It'd be pretty hard to spoof it. You'd need a large botnet of fake twitter users to beat the signal/noise ratio, and if they were geotagged you'd have to simulate an outward ripple from an epicenter. It's also pretty simple (trivial, if they've signed up to your app) to weight warning tweets by the user's "realness" rating, the number of inactive/fake followers, their location, that sort of thing.

Sounds like the same principle as Google Trends

[Meshach]

Google Trends looks at when the number of people searching for a specific term spikes and the correlation with major news. Ted is not a new idea.

Re:Sounds like the same principle as Google Trends

[buchner.johannes]

Yes it is, because Google Trends only analyses the past; connecting it to something useful, like an alert system, is something new. Unless you go extreme, but then every idea is based on something, so no idea is really new.

Re:Sounds like the same principle as Google Trends

[Narcogen]

Google Trends accurately measures what it purports to: people using Google to search for something.

TED doesn't detect earthquakes. It detects people twittering about earthquakes, and assumes either that correlation equals causation, or that false tweets won't rise above the level of background noise.

The real question is... twenty minutes for other reports? We can't put a seismograph online in real time, or are we afraid to do so?

How about a free, open source, crowdsourced Kickstarted network of online jury-rigged seismographs, then?

Fail whale, anyone?

[multicoregeneral]

Twitter crashes when it's overloaded. Earthquakes, and events like it cause a high enough amount of traffic to bring twitter down. People's lives depend on earthquake information systems. What happens if there's an earthquake at the same time as a Kardashian sex video? Or a Politician saying something stupid at a convention? We're all screwed then.

I don't see much point in this

[Y-Crate]

It will only report quakes that are already over. News reports, online reporting "Did You Feel It?" pages, etc already do a pretty good job of telling seismologists that something just happened.

Valuable earthquake detection would be detecting the P-Wave from a quake in progress, and automatically broadcasting a SAME Code, combined with some kind of equivilent forcibly pushed to every cell phone connected to a tower. Japan has something like this already. California is kinda, sorta working on it, but I'm pretty sure it's grossly underfunded and not really a priority.

Earthquake models suggest a quake on the northern or southern reaches of the San Andreas fault would reach Los Angeles in about 40 seconds. That's actually a huge chunk of time.

Let's assume:

- 20 seconds to detect a quake / automatically crosscheck with multiple sensors and transmit a warning to a predefined area.

- 5 to 10 seconds for devices to receive, decode and go into alert mode. Weather radios are always listening for SAME transmissions and can decode more or less instantly (assuming the user has programmed in their location). Cell networks could probably get the data there in the time it takes for a regular text message to arrive.

- That gives you 10 to 15 seconds to pull your car over, stop doing delicate surgery, stop fixing your roof, etc and find something to crawl under. It also gives you time to trigger automated fail safes. Gas valves can be set to close, emergency generators can be spun-up, fire pumps can activate, elevators can go to their recall floors and hold their doors open, while fire station doors can roll-up on their own and lock in place.

Re:I don't see much point in this

[zalas]

I think Twitter is a bit faster than that. Twitter users in Japan seem to respond really fast when they feel any moderate level of shaking; at times, if you follow enough Japanese people on Twitter, your entire timeline gets filled with people saying "oh hey, something's shaking" or "it's rocking" or "boobs!". So yes, you will get advanced warning if there are people closer to the epicentre than you posting on Twitter (and as long as they are not using a certain phone provider which got overloaded during the big earthquake/tsunami last year while all the other providers were fine). And yes, this obviously doesn't work if the earthquake knocks out all the cell phone infrastructure in the areas between you and the epicentre.

Amusingly enough, I was watching a UStream broadcast run by some Japanese guy and people from a different area of Japan told him that they just had an earthquake, and he replied saying he didn't get any info here. Then, several seconds later, the earthquake alarm went off in the broadcast. So, I think Twitter isn't going to be very far off in terms of speed and definitely should be able to inform you about an ongoing earthquake as long as it's not a super short one and you're like next to the epicentre.

Re:I don't see much point in this

[Y-Crate]

I think Twitter is a bit faster than that. Twitter users in Japan seem to respond really fast when they feel any moderate level of shaking; at times, if you follow enough Japanese people on Twitter, your entire timeline gets filled with people saying "oh hey, something's shaking" or "it's rocking" or "boobs!". So yes, you will get advanced warning if there are people closer to the epicentre than you posting on Twitter (and as long as they are not using a certain phone provider which got overloaded during the big earthquake/tsunami last year while all the other providers were fine).

I really can't believe that the reacting, tweeting, flagging, aggregating and alerting could happen in 20 seconds or less. Not a chance.

I find it extremely unlikely that tweets you're referring to were posted within 1/3 of a minute of the first shaking taking place.

Not to mention, that all of that aside, you still need a representative sample of people from a geographic area with public Twitter accounts. Unless you're OK with 8 people from the Salton Sea triggering an automated alert chain all the way to Santa Barbara. And you have to make sure no IP spoofing is going on to trigger a fake alert.

And then you have to make sure they're not dummy accounts. I don't have to tell you Twitter is absolutely filled with them.

"Sorry 3.8 million residents of L.A. who are without hot water / ability to cook until SoCal gas can come out and reopen all of your valves - a few people nonexistent people tweeted about an earthquake."

And then you still have no way to quantify the intensity of the shaking. A 5.5 is the same as a 7.2 in the minds of a Twitter user watching things fall from their shelves during the first 20 seconds of a quake.

If you don't have a rough approximation of how intense the quake is, you have no way to select the appropriate response.

This whole concept screams of another "WOW WEB 2.0 CROWDSOURCED USER CREATED CONTENT NONTRADITIONAL MEDIA" thing that moves forward based on the inertia of its own perceived specialness.

Re:prone to

[plover]

It'd be pretty hard to spoof it..

Not that long ago, 4chan would have done it for teh lulz. Statements like that are to them like a red flag is to a bull. However, they're all butthurt now because of the crackdowns on Anonymous, so I don't think they'd mess too much with a government agency anymore.

Automotive analogy

[aNonnyMouseCowered]

There's a car crash. The neighbors scream. By hearing the neighbors scream and listening to their remarks, I conclude that there's been a car crash. This works well so long as the neighbors aren't practically jokers.

Sure Twitter's a much bigger neighborhood and more immune as a whole to pranksters. But what if organized mayhem like Anonymous decide to pull off an out-of-season April Fools Joke and start tweeting #alieninvasion? There goes your Armageddon Detection System.

Japan

[fullback]

There is a system in Japan where we get a flash of information on TV even before we feel a strong earthquake. I had a few seconds to brace before many of the hundreds of aftershocks after the large 9 Magnitude quake last March 11. Obviously, the warning time depends on how far you are from the epicenter.

Re:prone to

[outsider007]

Since when is Twitter a government agency?

what if the internet is down?

[Osgeld]

two birds shit on the wire ours goes down, cant imagine if a serious earthquake happens

News flash

[viperidaenz]

Brand new system is faster than legacy system!

Re:prone to

[Neil Boekend]

Ah, when one has the future sight it's easy to mistake past for future. Tenses get a bitch if that happens. It can also mess up conversations.
medium: Yes I can.
normal: Is it true you can see into the future?
Plover was probably living in a Firefly-like future and posting in the present time.
Remember: there is no future but what we make for ourselves.

Courtesy of Terry Pratchett, Joss Whedon and James Cameron.

'May take up to...'

[sociocapitalist]

"Tweets in this case, can be picked up faster by researchers as compared to scientific alerts that may take up to 20 minutes."

First, 'up to' is bullshit that is useful in getting mobs to buy stuff but doesn't actually mean much at all.

From TFA:
"Paul Caruso, from the US Geological Survery, said: 'We do have sensors and it usually takes about five minutes before the sensors will see the earthquake."

So already it's five minutes and not twenty.

Second, it seems that the scientists have a problem with their seismology equipment if such equipment takes five minutes to register a (7.6 magnitude in this instance) quake.

Appears straightforward

[rotenberry]

The method described here seems like a straightforward idea.

Wasn't a similar idea posted to slashdot five years ago?

http://slashdot.org/comments.pl?sid=440258&cid=22283136