Rescuers question what3words' use in emergencies
bbc.co.uk
bbc.co.uk
https://www.addressingtheunaddressed.org/ . If you want to use this tech + operations expertise in other geographies...just reach out to them. They intend to share their knowhow and expertise.
https://www.youtube.com/watch?v=Hd3gGspgVGs
Also from what i understand, after the success in Kolkata, this was then used in Native American communities.
P.S. Plus Codes are open source - https://github.com/google/open-location-code And they are well researched.
>The character set for Open Location Code was selected out of over eight billion possibilities, using a word list of 10,000 words from 30 languages. All possible sets were scored on whether they could spell the test words, and the most promising sets evaluated by hand.
>Plus codes can be encoded and decoded offline.
>Plus codes do not depend on any infrastructure, and so are not dependent on any organisation or company for their continued existence or usage.
Thanks for the link though, hadn't heard of plus codes.
https://github.com/google/open-location-code/wiki/Evaluation...
Making it clear when something is incorrect for a given context is quite a helpful feature. But you need to have the necessary context.
You wrote "lightbox" and I read "mapbox", I thought someone implemented an image carousel by adding photos instead of map tiles in a Google Maps-esque UI (like this but with unconnected pictures: https://www.nightcity.io)
It feels like it was a quick hack of base64-ish encoding of the long/lat that evolved into a product to get an internal promotion
Can't help but feel that this is a bad idea. From writing code, we know that the more code you use, the higher the chance of bug entering the code.
Also felt that the authorities are trying to extend the UK postcode system to the rest of the world.
https://en.wikipedia.org/wiki/Postcodes_in_the_United_Kingdo...
this is widely in use in the ham radio circle and it works great--if you know phonetic alphabets, it's easily understood over the air, truncation is well-behaved, and so on and so forth.
edit: for SAR you probably need 8 characters not the customary 4/6 character version
> duties.factory.person was located in China
I get Switzerland [1]
> refuse.housework.housebound was in Australia
I get Belarus [2]
> demand.heave.surprise was actually in Canada
I get China [3]
Now that's really confusing. Is that a bug in the system that changes the locations of three-word-codes, or is that (very) poor reporting?
[1] https://what3words.com/duties.factory.person [2] https://what3words.com/refuse.housework.housebound [3] https://what3words.com/demand.heave.surprise
> So, for example, circle.goal.leader and circle.goal.leaders are less than 1.2 miles (2km) apart along the River Thames.
The second one is in South Dakota for me: https://what3words.com/circle.goal.leaders
Edit: scratch that - it doesn't really lead to this problem does it.
Around midnight, we called it quits after standing in the middle of a major highway holdings a large metal plate with mag mount on it.
Hooked up to a gps that only displays lat long.
Fundamentally this is nothing different than say a base 10 to base 32 conversion as an algorithm and while its possible to mess something like that up if you are amateur enough it would show a huge level of technical incompetence.
Generally its a sign of a company that spends money and thoughts on marketing than on their product.
It's also not well designed, as plurals and homophones are not accounted for.
Avoid using or promoting this system.
I agree that it's not well designed and shouldn't be used.
TFA seems to suggest that this isn't such a list.
But if you take a look at the ripped off algorithm (google what free words, take a look at the html and the script block inside it), you'll see it includes an array of words with a specific order.
That list and the order the words are in are copyright protected. Copyright is about protecting the particular form of something. This a very specific word list in a very specific order (the indices matter). It's a clear cut case of those specific words in that specific order being protected under copyright. If you want to be compatible with w3w, you have to use that specific list.
If you write a book, poem, or article that's clearly a copyrightable work even if all of the words are commonly used ones. This not different in any meaningful way from the point of view of the law. The fact that the words in that order spell out nonsense is not relevant. Copying and using it in its entirety is not fair use and a clear infringement.
You could use the same algorithm with a different word list and then you'd merely be facing patent infringement here. I would not recommend building a business around such a thing until the patent expires.
I think this case further down is much more concerning:
> He found that the algorithm behind W3W often gave similar-sounding words and plural versions of words for locations in close proximity, which could cause confusion.
> So, for example, circle.goal.leader and circle.goal.leaders are less than 1.2 miles (2km) apart along the River Thames.
I wonder if the algorithm could somehow make it so similar sounding words like this aren’t anywhere near each other physically. It’s not as big a problem that leader and leaders sound the same, if you guarantee that only one of them is in the UK or at least within the same city/national park etc.
"So, for example, circle.goal.leader and circle.goal.leaders are less than 1.2 miles (2km) apart along the River Thames."
That's potentially catastrophic, without an easy way to spot the mistake.
Why What3Words is not suitable for safety critical applications - https://news.ycombinator.com/item?id=27058271 - May 2021 (35 comments)
Why What3Words is not suitable for safety critical applications - https://news.ycombinator.com/item?id=27029706 - May 2021 (3 comments)
What3Words: App used by emergency services under scrutiny - https://news.ycombinator.com/item?id=26985759 - April 2021 (14 comments)
by relying on context you can lower the area you need to have address combinations for. In practice you would simply repeat the addresses every couple of 1000km or so.
needing help and calling in rescue from the other side of the world while also being unable to say you are in such and such area is presumably not something that happens too often.
It strikes me that W3W will eventually be replaced by the emergency location standards built into mobile which effectively SMS your location to the control centre on starting the call. Then there's little need for voice based W3W exchange - at best it could be a dialogue between caller and call handler to confirm the approximate area and location, and to take details like (for urban environments) if it's an apartment block or house etc. Or off the beaten track, which side of the mountain route they walked up to reach their current location.
It would've been more logical to automatically share it, to be honest...
Person: "emm, I'm lost on the Yorkshire Moors...".
Rescuer: "no problem sir, just take advantage of the completely flawless wireless telephony system we have to download this app..."
Person: "ok, it says I'm at correct house battery"
Rescuer: "hmm, I'm showing you at the top of a mountain in Nepal. Let's try that again..."
If correct horse battery is in England and correct house battery is in China and you have a well designed lookup it would know you're looking for something in England.
correct house battery is in China at x, y. Do you mean correct horse battery in England at z, w?
The real danger comes when the mistakes are a few miles apart.
That said, having familiar words in your (local?) alphabet is charming. But for public good it HAS to be public domain (the alphabet + algo, not the service, however).
Not sure what's the average number of characters required for a 20m to 100m precision with such method.
It's true that the idea of plus codes of using city names makes a lot of sense.
Use numbers.
They need a way to transfer location data by the word of mouth, which is extremely difficult when you're using numbers. With longer numbers, such as a comparatively precise location, you have a high risk of errors on both ends, so searching for a solution makes absolute sense.
Nevertheless, based on the proprietary IP maybe the rescue services should consider building an own solution or evaluating open source alternatives if there are any
My bank get me to key in my sort code & account number while using telephone banking. If we did the same here, we could get the precision of GPS without the user needing to call out any numbers
It's not clear to me how a stupid system like this is used in favour of, say, 5 numbers+5 numbers (4 decimals coordinates lead to a ~5.5m radius)
I'm not sure I understand what you mean with "high risk of errors on both ends", as there is no chance of mispelling. How do I know that? Because numbers are used in military standards for comms that have to be clear even through unreliable channels. Languages that have numbers which might be misunderstood specifically use a different number set (one example is the military in China).
Building your own solution when you have a GLOBALLY ACCEPTED standard that took YEARS to be agreed upon is criminal in my opinion, rescue missions on the Alps where different cultures meet being a great example why you should stick to numbers.
In rescue we are talking about untrained people how give an initial communication by word of mouth. This data is then translated to coordinates and then communication by then trained people in the chain of rescue.
So the problem is this initial communication by untrained people.
Now add in hypothermia, noise, weather, or bodily harm to that equation.
The only improvement would be if the app could provide the NATO phonetic alphabet version of your coordinates to you - but you're not going to remember those without looking at your screen while talking - which is multitasking in a situation that you probably aren't going to do it well in.
Finally, the verbal transfer time is about 1/3.
I don't like this thing for the reasons provided (proprietary and clearly has issues), but would rather see this in use than trying to provide ten numbers while bleeding out.
regardless, this is an article about a UK company's product being used by UK emergency services. An injured foreigner who can't speak English is going to have trouble with numbers too (granted, you may learn numbers first). You don't just make a call to emergency services and start spewing out numbers with no other vocabulary.
I can see your point on an international scale, and agree with you that this product is not globally scalable. Language barriers are language barriers until we all have the fictional babel fish.
edit: besides, you can always provide phonetic 'translations' if you want in to a foreigner's language/script if you must - not my preferred solution either though.
That doesn't sound correct. Numbers are universal, easier to remember, and less ambiguous than three English words. I don't see how it's more error prone to use numbers based systems, which many organisations and communities have been using for decades. And also considering the length of the three words which is roughly the same.
Although the numeric location gets more precise as you add more numbers - if they only get 3 decimal places, that's about 110x80m in the UK.
If you use degrees, minutes and seconds, you only need 6 numbers to get ~30x20m in the UK (and a decimal place on the seconds brings it down to ~3x2m - the same as W3W.)
Weren’t there a bunch of people trying to do exactly that (FOSS, translatable etc) only to get sued or as-good-as by what3words?
There's a lot of research on how to choose words to ensure reliability of communication and reduce risk of misunderstandings, for example, the criteria used for Aviation Phonetic Alphabet - and what3words fails to take all that into account.
You don't want separate things that are pronounced nearly the same, what3words fails in this regard as it contains many pairs like wants/once, secretary/secretory, ordinance/ordnance.
You'd want a person that can reconfirm that word multiple to be certain that they've got that right, but what3words includes singular and plural wordforms (e.g. leader/leaders) which has caused miscommunication even when the responder has understood that they've got the right concept for the word or when the person who recites the three words from memory mistakenly thinks they've memorized it right.
You'd want to ensure that the words are limited to a set of sounds that are clearly pronouncable and understandable by non-native speakers of different backgrounds, what3words does not even attempt that.
In short, the particular system is poorly designed and fundamentally flawed because of that.
Not sure it's good enough over a crappy phone line...
With a geohash someone only needs to provide 7 to 9 characters depending where they are and the accuracy needed.
Also as each additional character in the hash adds accuracy it should be easy for an operator to easily sanity check i.e caller to operator "I am hiking in the lake district...my geohash is ..." the operator only needs the first four characters to confirm if the general location is correct or not.
I guess it's still prone to spelling errors from operators inputting the data from a call.
As others say if you are installing an app to give you a location i.e. a geohash you can use an app to give a lat/lng.
You can trivially do distance calculation by using their algorithm (which works offline if you use their sdk) to get the coordinates. The DB space is completely irrelevant: this algorithm is intended for humans being able to memorize three words. So, not sure what you are going on about here. Again, they have paying customers that are getting value out of this. Hard to argue with that, which you seem to be attempting here.
I'm not aware of a better solution that is usable by normal people. Good luck communicating a geohash to the emergency services. It's not going to be a thing as much as I like using them. Placekey seems one of their competitors but they don't produce human readable codes and they are US only. In any case, most people don't even know their latitude or longitude to any kind of precision. Coordinates are hard for people.
Regarding the IP, they use a combination of patents (the algorithm) and copyright (the word list) to protect their IP. They built a customer base around this stuff and they are well within their rights to protect their own work, which is what they are doing. IP law is what it is and w3w is simply using that as it was intended (protecting IP). That's just normal behavior for VC funded startups. Try to get some IP through some honest work, protect it, build a business around it. They are not being patent trolls here.
The way I see it, w3w has built their business without edge cases like in this article being a major obstacle to them. You could argue emergency services are important enough that they might want to improve their word list a bit. But still, it's not a basis for dismissing their business model, solution and general right to exist.
IMHO they are a bit over valued at this point and a likely acquisition target for someone so the investors get their bail out. Their current strategy is a long term dead end from the point of view of people adopting this. The proprietary nature of their solution is an obstacle for that. Once patents expire, similar solutions might emerge with different word lists, numbers of words, etc. that are more successful. Alternatively, post acquihire, a visionary company might just use some more liberal licensing to make this a bit more attractive.
Moving to four words could really help remove the need for plural and words that sound similar. Accents really do make things tough.
Their business now involves lots of bizdev droids and threatening open source implementors with baseless copyright lawsuits. It's nice work if you can get it.
I also feel like the solution W3W is trying to deliver is a short term patch until we get emergency SMS location from phones themselves delivered to the control room - having in band spoken location works until you have that, but W3W leaves a lot to be desired. Given it is an app, it might be easier for it to show the user words to read, which are themselves all just NATO phonetic alphabet words to spell their location. You'll get far more redundancy and human error correction over the voice channel using the phonetic alphabet than words.
It wouldn't be as shiny a solution though, and that probably doesn't play into W3W's marketing strategy... Even if it would help save more lives.
Not a good system.
At least wiki brightened my mood : The site has been parodied by others who have created services including What3Emojis[29] using emojis, What3Birds[30] using British birds, What3fucks[31] using swear words and What3Numbers[32] using OpenStreetMap tile identifiers.
The only justification is the ease of communication over the phone.
I only don't like that it's proprietary and the company responsible is pretty aggressive about enforcing their rights - there shouldn't be a place for that in public emergency services.
I mean.. as opposed to two numbers? A slow speed high error correction audio encoding of them would be easy. Call emergency services, press the button, they decode the tones, immediate, accurate, simple.
If you mixed the calling and playback into a single action, you could make calls and direct people without even the ability to speak.
Maybe it helps with an issue they have noticed with certain callers that I'm not aware of.
In situation A, it's going to be a lot quicker and clearer to communicate w3w than an ll pair (for the same precision a ll pair requires 12-18 digits, plus specificity on +/-).
In situation B it's going to be a lot easier to remember the w3w descriptor, and you still have the same advantage as in situation A.
For spoken communication, w3w's rival is "it's 123 Fake Street, I'm the bloke in the green parka", not ll pairs.
So they knew exactly that it was wrong. So it worked perfectly.
If you want to be hip you could write a program to work out what was actually said given other information, ie what cell tower it came from.
I hate 3 words, I can't remember why, maybe it was copyright or something, but this article isn't selling why it's a problem.
Rescuers will also get all the normal info like the parents who might know the track, and everything else said.
Saying just use latitude and longitude co-ordinates, is like saying "MREW is keen to encourage anyone planning a day in the outdoors to carry a paper map and a compass"
Latitude and longitude is scary and complex. Rescuers might think it's easy but I'm sure there's plenty of things they find scary and hard other people find easy. Just the appearance of most of the latitude and longitude apps is technical. 'What 3 Words' sounds like teddy bears.
Lat/Lon is already integrated in all mapping apps or GPS devices.
You can easily make a lat/lon visualizer with simple html and JavaScript; it would require GPS and a web connection, but so does this stupid what3words.
> So they knew exactly that it was wrong. So it worked perfectly.
Okay, but what if they got the invalid location and then the caller disconnected? Calling them back might be difficult. And if you do get the caller to repeat, can you now be sure that "jump.legends.warbler", which is in the UK, wasn't supposed to be "jump.legends.warblers", which is also in the UK? And how many cases are there that they didn’t mention, which were in the correct country, but a few kilometers off the real location?