Great!, codes are shorter than on ubicate.me, the main difference is that Location Aware Names try to use the encoding as names , did you see Location Aware Street Names on slide 44?
My humble advice for this slide deck is: get to the demo. In this case, it's phonetically encoded coordinates. It's a cool idea, but requires adoption of a "majors system" like system for converting numbers into sounds. That's asking a lot!
Thanks for the advice michaelq, I will change the slide to get to the rel life apps first and explain the concept later!.
What do you mean by "major system"?, could you explain more?.
The idea is to use what we already have now, "Names" but encoding information inside them, if you get to the Location Aware Street Names slides, you can see how it can be used to make some very interesting stuff, fast forward to slide 55.
Also when you publish a presentation it is really better off just give the non-animated version (even the pdf version with 4 pages to just load from pic 1 - pic 4 is still not a good idea).
I made a non animated and "cut to the chase" slideshow you can acces on a "Slides are too slow?" link now on the website, hope that improves the experience!.
Seems like the easiest metaphor for this is "DNS for GPS" - i.e. in the same way DNS translates IP addresses into something human-readable, this translates lat-long into something easy to remember and work with.
I disagree. No algorithm connects google.com to whatever IP Address it is associated with.
In this case, there is a one-to-one correlation. Functionally it is somewhat similar, but you could figure out the location of one of these crazy names by running through a simple algorithm, locally. You don't need to make requests to a remote server, to do the DNS lookup.
I would also disagree that these "names" are easy to remember. :) They are pretty random. I doubt these names follow any of the "rules" that human language does, either. They are mapped to a system of numbers. Human languages share some common rules and organization, I believe.
I'm not sure what you mean by applying the algorithm to IP quads? If you give me the name "google.com", there is no algorithm I can apply which yields the IP address, correct? Because that IP address could be reassigned to a different name, right? The only way I can make the connection is to do a DNS lookup, which I thought is essentially just searching a database for "google.com" and returning me the corresponding IP address.
I like what3words - it's about as good as you can get for the same goofy idea. Those ubicate names are shorter but ... ugly. I like that my data centre is at *fish.wages.jelly but not TACEGE-CARAFAN.
If I'd heard my office name over the phone using the ubicate system I might instead head for http://ubicate.me/TACEGE-CARAVAN ... which looks to be about 30 miles off the coast of Holland, not on the York ring road :-)
So I get the idea, but this algorithm produces awful names.
what3words noticed that you needs less than 1800 words to assign a combination of 3 words to every square meter on earth, so those words can be chosen to be different from each other, unambiguous, common etc. Still ENGLISH and not very international, but it's a much better idea.
They do support a lot of other languages, although the words aren't the same between languages as far as I can tell (so they can't be 'translated'). Still, I bet they cover most of the world's population in their local language.
Totally agree, What3words is way better for communication. But even there, if you read me the locator, I have no idea where the location is. With traditional addresses, I have a good idea of where things are, and how I might get there, just by hearing the address, the same is true with lat/lon or MGRS to a lesser extent. This system means I need a device to figure out where anything is.
This whole idea seems like a variation on "the world would be better if only it was run by computer logic" (where better = more orderly and machine readable). For whatever reason, other considerations (cultural affinity, convenience in a daily context, etc.) seem to be more important to most people.
The slides are really cool – there's a lot of depth. I was really impressed that vectors could be encapsulated in names, removing the need for street information to be stored. Can the system encode non-linear vectors, i.e. curvy streets?
Thanks rusew! , at the time I was able to encode points and street names (lines) , hopefully more complex non-linear vector can be also encoded into short name forms.
New York is confusing, as York is a real place? :)
Tongue in cheek of course, but I don't think that _that_ problem would limit the adoption. I fail to see how this might reach a critical mass, but ambigious (parts) of names? We already have that today and handle that fine in general.
Its a nice little thought experiment, but current place names have been optimized for human memory. Objects, names, events are all synaptically linked to enable the average human to master a relatively complex set of locations. The provided system also completely excludes cultural background (also important for how "good" a location name is for a given person). tl;dr: Systems like this already exist (e.g. longitude/lattitude). The proposed system is on the same usability scale for individuals like existing systems. Does not improve anything.
Thanks for your comments MrDosu, I definitely agree with you, but the Location Aware Names concept can be used to take language and cultural factors into account , to create a reference that is indistinguible from any other non Location Aware Name , fast forward to slide 47 and let me know if that answer your comments!
Why would a town that historically has not needed street names have a need for convulated "street ids"? Small towns like this dont have street names because they are superfluous. Mail arrives, people find places.
"50% of people in developing countries live in places where there are no maps, addresses or street names, UBI gives them an immediate alternative for their needs."
Because everybody's got a device that has GPS and can decode these opaque strings, right? /facepalm
GPS coordinates work regardless if there is an image on the Google map... a coordinate is a coordinate (which is why the military, aviation industry, maritime industry, agricultural industry, etc... all use coordinates instead of some silly system like this phonetical one.)
Plus, do those 50% of the people tend to get lost going from place to place in their daily lives? And do members of the other 50% have a need to be able to precisely locate any of those people without maps, addresses, or street names?
I humbly submit that this is a very nice solution in desperate and ultimately fruitless search for a problem.
You would still need another line for multi-story buildings. For example, I live on the 2nd floor of a building. On google maps, my flat is located above a restaurant. I guess I would have to write "apartment 3" before the ubicate address.
Neat idea, but things that become really successful slipstream into the existing way of doing things, rarely do they supplant it in such a radical way. If the US can not even change to the metric system, think how hard this will be to adopt.
Maybe a chance in third world countries, but then you need to partner with all major online map makers to make this the default or a prominent identifier in all their softwares.
Place names have huge emotional content associated with them. Place names are changed for example to honor a person (Washinton, DC) or an important concept (Union City, CA). There's something very human and basic about that mapping between place names and our identities.
I think this is an innovative idea you propose, but because it runs against the grain of what it means to be human, it may not receive the uptake you would like.
I certainly agree with you, but you dont have to replace the names you are already using to take advantage of this, you can say:
not_that_noob
Washington Ave 123
Washington DC
BIFAPU-BATEDU
Floods, tornadoes, tsunamis and natural disasters can wipe out landmarks and leave addresses or conventional signaling unusable , that last line can make all the difference in such situations since it remains always unalterable.
And then you have Location Aware Street Names, where you can use words in your language to form street names, so you might use words that are more cultural meaningful to us.
I don't think I'm on board with this. Here are two names from the same street:
NOZAXO-JEJEZU, NUZAXOJ-JAJEZUN and here's one from another location in the city PABAXO-GAKEZU. First, I wouldn't like to prominently use these names on my business as I would be naming them nearly the same as any other one on my street. Second, it is not clear at a glance that the second address is in the same area. And without specifically saying "Atlanta" you lose all types of information about taxes, police, fire and other government type things which apply to areas bound by complex curves in lat-lon space.
Here's three other addresses:
BAJASU-BONEXO, DAFASU-DUTEXON, LIDEJA-KESERO
Two of these are about 10 miles apart and the other is thousands of miles away. I can see some similarities to the ones close by but if just given a single UBI name, I doubt I'd have any clue where it is. It seems that actually seeing lat-lon numbers might make it easier to see that locations are close together.
This is also ignoring streets, which carve complex paths through lat-lon space. I'm not sure if the street name is wrapped up in these UBI names but losing that information would be terrible for locals or honestly anyone else traveling. If you know something is on "broad st." and you are on broad st., then you can follow the road to get there. If you are looking for NUZAXOJ-JAJEZUN and you are at NOZAXO-JEJEZU, you may not realize that you are on the same street.
I didn't go through the entire set of slides as it was pretty onerous, some having 7 clicks to reveal everything in them. I think I got the basic idea though and I don't really see a use case for this replacing any part of our current system. I would never name a business using these. They are unnatural and difficult to pronounce, so I'm not sure how the "Pine Hill" example is valid. Adding these to existing addresses would be useful, in the sense that adding lat-lon to an existing address would be useful but I don't find them more readable than lat-lon. And the UBIs have the negative effect of needing to be translated to lat-lon in order to find locations.
That is the big take-home problem here. I looked up the address that they said was in Nigeria (it's actually in Mali), REJATE-POKAJA. I wanted to know, "well, what's the address of the building across the road?" and it turns out that that is: RIJATE-PIKAJAS. I think my first thought is, "Is that really what you want?" Those sound pretty darn similar, and it'd be easy to get confused. In reality you already have a lot of info about a place implicitly, so you want to focus on the stuff that you don't know, magnifying that to cover the whole number.
It seems that they get something like 16-20 bits per word or so. I think my first approach in this case would be to feed lat/long through a cylindrical equal-area projection, since the math for those is pretty simple (so it can be implemented in diverse systems easily) and at least there is a semblance of fairness to the resulting numbers -- but really, I think what you want is to cover the Earth with a massive geodesic dome of points and just number the vertices with some algorithm, perhaps placing a great circle on the Greenwich line and then descending around in some sort of spiral. There's no fundamental reason I can see why we'd need to project at all, really.
So, you form two numbers with the Gall-Peters projection:
2^9 * (most significant 9 bits of x) + most significant 9 bits of y
2^9 * (least significant 9 bits of x) + least significant 9 bits of y
or else you label each of 2^18 geodesic faces in a spiral and do coordinates on the faces; either way.
You take these numbers, and multiply them by some large prime, modulo the number that you've got. Prime multiplication is a simple way to permute the numbers so that adjacent vertices are unlikely to be numbered similarly. If you really wanted, you could then XOR it with some fixed pattern and multiply by another large prime, which may help with names being similar at the poles themselves or something (depending on the size of the prime). Any quasi-random permutation will work, I just figure "let's take the ones that are only a couple fixed lines of code." The nice thing about these permutations, too, is that they're invertible: someone gives you a name, you convert it to a number, you pass it back through this, and you get map/geodesic coordinates.
Then you convert those numbers into syllables. This way you can know that your city is mostly DAFAXO except the northeastern part is part of LINUZU, completely different names. The most significant bits are thus scrambled but isolated. The least significant bits are scrambled separately and give every point on the ground a wildly different name, so that nobody confuses your place with one down the street.
Hi Drostie,
It is an excellent analysis, when designing the algorithm there are 2 choices, to make the Global Addresses somehow related, so you can know that REJATE-POKAJA is near RIJATE-PIKAJAS, or to make them totally different so you wont get confused by the similarity, both approaches have their advantages and disadvantages , and as you say, is not something difficult to implement
Hi Wuliwong and thanks for your comments is a great feedback, Im Robert the developer behind the idea.
Location Aware Names used as Global Addresses is one of the choices you can use. If you want to keep the current Street & Number format you can use Location Aware Street Names, so your business address can be at street "Clerigo Tigorros 155" and the building in front of you will be at "Clerigo Tigorros 158", you can use it just like any other street with the Street name and number.
You can get there using your offline address like "Clerigo Tigorros 155@gazado-fifene" or even if the town is already on a map using "Clerigo Tigorros 155@(Naranjito, itapua, Paraguay)" so every street in town will share the same Zone code @gazado-fifene if you want to know if they are in the same area.
In case you wonder this name was generated for a spanish speaking town, clerigo means "cleric", so you can use any language you want to generate the street names.
I made a non animated and "cut to the chase" slideshow you can acces on a "Slides are too slow?" link now on the website, hope that improves the experience!.
If you have any other question or suggestion let me know.
I'm concerned my street and the street next to me are very similarly named. I suppose this can be confusing in the long run as similar places will have similar names.
it might work better as a more chaotic (two-way) hash function, true - next door streets different by at least half the letters, or something. Or at least with a prefix for the higher-significance bits, like "town", and a suffix for the lower significance bits, like "street".
How resilient is this type of encoding? i.e. if I pronounce this to someone over the phone, they may write down the phonetics incorrectly or with a different take on the spelling.
What happens if they do that?
I see a lot of names depending on consonant-E and consonant-A pairs, but in the English language how you pronounce those sounds can wind up being very similar dependent on context.
Yes, I absolutely agree. Several of the addresses I encoded would be easy to mistake via voice which is why I thought the "Pine Hill" example very poor.
The example phonetic address they use for a bunch of examples is "RERI-NUCA". Not very memorable.
And what happens if you misspell it? The example of spelling "Pine Hill" as "Pine Hil" is easy for a human to see and correct; would I be able to realize that "RERY-NOOCA" was a typo?
These are also utterly useless for navigating without GPS. Even if we assume that all maps are replaced with ones that have this letter-based coordinate system as their grid, I still can't see it being easy to give directions, or to navigate without a phone.
How does this work for languages that do not use Roman characters?
There is a scheme for creating "street names" in this (though it doesn't seem to handle streets that turn). However these names are all pretty much gibberish names. And naming all streets this way would remove the local character of place names - a Los Angeles street name has a different character from a New Orleans street name, and a Parisian street name has a stranger one yet.
I dunno. It's moderately clever but it doesn't really feel like it's designed for human-scale interactions.
This is yet another attempt[1][2] to turn the already-very-good-and-precise coordinate system into something more overly-complex and unnecessary.
If you need to meet at exact locations where both parties don't have a common name for and no commonly-assigned address exists (like someplace in the middle of a large park), just provide a coordinate -- it will be accurate and all GPS/Map systems can take them, including your phone.
100% agreed. In Atlanta, we name all our streets some version of "Peachtree". Definitely gives a feel to the place. Doesn't help navigating so much, but whatevs. :) In Philadelphia, we have an area called the "tree streets," for Chestnut St., Walnut St., etc. Just can't see that stuff going away.
Oh yes, and the Roman characters. I was thinking about that. I'm pretty sure in China, loads of popular websites are just number strings. I can't imagine that this UBI system would be better for them. So knock off 1 billion people right there. hah.
This is a rather wonderful concept. English names are good but often when names in another language are converted into English and typed into Google maps they can be spelled many different ways, .e.g. "Dwarka Mor" or "Dwarika Mod" or "Dwarika Mode" or 10 different other combinations (which Google maps using Google suggest can fix thankfully)
But how does this handle misspellings? Not sure if it already has it but there should be CRC check in the first or last letter to make sure the address printed misspelled a C instead of a G.
Great idea superasn, right now if the misspell result in an invalid address it will let you know, but you can make a mistake that result in a valid address anyway.
107 comments
[ 3.6 ms ] story [ 94.2 ms ] threadThe idea is to use what we already have now, "Names" but encoding information inside them, if you get to the Location Aware Street Names slides, you can see how it can be used to make some very interesting stuff, fast forward to slide 55.
Just my 2 cents.
In this case, there is a one-to-one correlation. Functionally it is somewhat similar, but you could figure out the location of one of these crazy names by running through a simple algorithm, locally. You don't need to make requests to a remote server, to do the DNS lookup.
I would also disagree that these "names" are easy to remember. :) They are pretty random. I doubt these names follow any of the "rules" that human language does, either. They are mapped to a system of numbers. Human languages share some common rules and organization, I believe.
And how does one achieve the sort of critical mass where the average person on the street knows their own home location?
If I'd heard my office name over the phone using the ubicate system I might instead head for http://ubicate.me/TACEGE-CARAVAN ... which looks to be about 30 miles off the coast of Holland, not on the York ring road :-)
So I get the idea, but this algorithm produces awful names.
what3words noticed that you needs less than 1800 words to assign a combination of 3 words to every square meter on earth, so those words can be chosen to be different from each other, unambiguous, common etc. Still ENGLISH and not very international, but it's a much better idea.
This whole idea seems like a variation on "the world would be better if only it was run by computer logic" (where better = more orderly and machine readable). For whatever reason, other considerations (cultural affinity, convenience in a daily context, etc.) seem to be more important to most people.
Tongue in cheek of course, but I don't think that _that_ problem would limit the adoption. I fail to see how this might reach a critical mass, but ambigious (parts) of names? We already have that today and handle that fine in general.
"I found my thrill / on MARAMUT-GACIRA"
"JODEGE-SIFAJAR is in my ears and in my eyes"
"JACEGE-RURAJA forever"
Because everybody's got a device that has GPS and can decode these opaque strings, right? /facepalm
I humbly submit that this is a very nice solution in desperate and ultimately fruitless search for a problem.
Maybe a chance in third world countries, but then you need to partner with all major online map makers to make this the default or a prominent identifier in all their softwares.
I think this is an innovative idea you propose, but because it runs against the grain of what it means to be human, it may not receive the uptake you would like.
I certainly agree with you, but you dont have to replace the names you are already using to take advantage of this, you can say:
Floods, tornadoes, tsunamis and natural disasters can wipe out landmarks and leave addresses or conventional signaling unusable , that last line can make all the difference in such situations since it remains always unalterable.And then you have Location Aware Street Names, where you can use words in your language to form street names, so you might use words that are more cultural meaningful to us.
NOZAXO-JEJEZU, NUZAXOJ-JAJEZUN and here's one from another location in the city PABAXO-GAKEZU. First, I wouldn't like to prominently use these names on my business as I would be naming them nearly the same as any other one on my street. Second, it is not clear at a glance that the second address is in the same area. And without specifically saying "Atlanta" you lose all types of information about taxes, police, fire and other government type things which apply to areas bound by complex curves in lat-lon space.
Here's three other addresses: BAJASU-BONEXO, DAFASU-DUTEXON, LIDEJA-KESERO
Two of these are about 10 miles apart and the other is thousands of miles away. I can see some similarities to the ones close by but if just given a single UBI name, I doubt I'd have any clue where it is. It seems that actually seeing lat-lon numbers might make it easier to see that locations are close together.
This is also ignoring streets, which carve complex paths through lat-lon space. I'm not sure if the street name is wrapped up in these UBI names but losing that information would be terrible for locals or honestly anyone else traveling. If you know something is on "broad st." and you are on broad st., then you can follow the road to get there. If you are looking for NUZAXOJ-JAJEZUN and you are at NOZAXO-JEJEZU, you may not realize that you are on the same street.
I didn't go through the entire set of slides as it was pretty onerous, some having 7 clicks to reveal everything in them. I think I got the basic idea though and I don't really see a use case for this replacing any part of our current system. I would never name a business using these. They are unnatural and difficult to pronounce, so I'm not sure how the "Pine Hill" example is valid. Adding these to existing addresses would be useful, in the sense that adding lat-lon to an existing address would be useful but I don't find them more readable than lat-lon. And the UBIs have the negative effect of needing to be translated to lat-lon in order to find locations.
It seems that they get something like 16-20 bits per word or so. I think my first approach in this case would be to feed lat/long through a cylindrical equal-area projection, since the math for those is pretty simple (so it can be implemented in diverse systems easily) and at least there is a semblance of fairness to the resulting numbers -- but really, I think what you want is to cover the Earth with a massive geodesic dome of points and just number the vertices with some algorithm, perhaps placing a great circle on the Greenwich line and then descending around in some sort of spiral. There's no fundamental reason I can see why we'd need to project at all, really.
So, you form two numbers with the Gall-Peters projection:
or else you label each of 2^18 geodesic faces in a spiral and do coordinates on the faces; either way.You take these numbers, and multiply them by some large prime, modulo the number that you've got. Prime multiplication is a simple way to permute the numbers so that adjacent vertices are unlikely to be numbered similarly. If you really wanted, you could then XOR it with some fixed pattern and multiply by another large prime, which may help with names being similar at the poles themselves or something (depending on the size of the prime). Any quasi-random permutation will work, I just figure "let's take the ones that are only a couple fixed lines of code." The nice thing about these permutations, too, is that they're invertible: someone gives you a name, you convert it to a number, you pass it back through this, and you get map/geodesic coordinates.
Then you convert those numbers into syllables. This way you can know that your city is mostly DAFAXO except the northeastern part is part of LINUZU, completely different names. The most significant bits are thus scrambled but isolated. The least significant bits are scrambled separately and give every point on the ground a wildly different name, so that nobody confuses your place with one down the street.
Location Aware Names used as Global Addresses is one of the choices you can use. If you want to keep the current Street & Number format you can use Location Aware Street Names, so your business address can be at street "Clerigo Tigorros 155" and the building in front of you will be at "Clerigo Tigorros 158", you can use it just like any other street with the Street name and number.
You can get there using your offline address like "Clerigo Tigorros 155@gazado-fifene" or even if the town is already on a map using "Clerigo Tigorros 155@(Naranjito, itapua, Paraguay)" so every street in town will share the same Zone code @gazado-fifene if you want to know if they are in the same area.
http://ubicate.me/street/index?utf8=%E2%9C%93&street=Clerigo...
In case you wonder this name was generated for a spanish speaking town, clerigo means "cleric", so you can use any language you want to generate the street names.
I made a non animated and "cut to the chase" slideshow you can acces on a "Slides are too slow?" link now on the website, hope that improves the experience!.
If you have any other question or suggestion let me know.
- Nearby places have very simillar names; Just changes one letter "U" with "O"
What happens if they do that?
I see a lot of names depending on consonant-E and consonant-A pairs, but in the English language how you pronounce those sounds can wind up being very similar dependent on context.
And what happens if you misspell it? The example of spelling "Pine Hill" as "Pine Hil" is easy for a human to see and correct; would I be able to realize that "RERY-NOOCA" was a typo?
These are also utterly useless for navigating without GPS. Even if we assume that all maps are replaced with ones that have this letter-based coordinate system as their grid, I still can't see it being easy to give directions, or to navigate without a phone.
How does this work for languages that do not use Roman characters?
There is a scheme for creating "street names" in this (though it doesn't seem to handle streets that turn). However these names are all pretty much gibberish names. And naming all streets this way would remove the local character of place names - a Los Angeles street name has a different character from a New Orleans street name, and a Parisian street name has a stranger one yet.
I dunno. It's moderately clever but it doesn't really feel like it's designed for human-scale interactions.
If you need to meet at exact locations where both parties don't have a common name for and no commonly-assigned address exists (like someplace in the middle of a large park), just provide a coordinate -- it will be accurate and all GPS/Map systems can take them, including your phone.
[1] https://news.ycombinator.com/item?id=8052908
[2] http://www.mapcode.com/
But how does this handle misspellings? Not sure if it already has it but there should be CRC check in the first or last letter to make sure the address printed misspelled a C instead of a G.