9 ms·

The reminds me of people's attitudes towards reCAPTCHA - started by researchers at Carnegie Mellon University. "Stop spam, read books." Everyone was supportiv

by chippy 11y ago

The reminds me of people's attitudes towards reCAPTCHA - started by researchers at Carnegie Mellon University. "Stop spam, read books." Everyone was supportive of it when it was used towards the non profit digitization of out of copyright books. So, started externally, Google ran with it and continued and expanded the range of books. It remained good. Apparently loads of books were digitized. Then, possibly along with the change in this article - or along with the perceptible change within Google where everything had to be business accountable a few years back, reCAPTCHA started being used to digitize address numbers from StreetView to improve Googles online mapping, geocoding offering. Nothing to do with books, nothing to do with improving the world. Now reCAPTCHA is being used for image recognition and training (identify the images with salads). Nothing to do with books, information or improving the world - everything to do with Googles own offerings. What's even sadder is that http://captcha.net/ http://captcha.net/ still states that it is being used to "help digitize books" but all the links go to https://www.google.com/recaptcha/intro/index.html https://www.google.com/recaptcha/intro/index.html which have all but removed any public benefit wording. " "Stop spam, read books." was removed from Googles site in 2014.

blfr 11y ago
How is reliable Google Maps and Street View not improving the world? I'm using them daily and they're improving my life a lot.
- tdkl 11y ago
  At what cost?
  - golergka 11y ago
    At cost of solving the captcha problem?
- blub 11y ago
  They are improving your life in exchange for information about you and the places you visit. Their product is closed, the proprietary data is not accessible outside their apps, or if they wish to license it. OpenStreetMap is improving the world.
  - mrb 11y ago
    "the proprietary data is not accessible outside their apps" This is not true. Many APIs are available, specifically to allow building 3rd-party apps: Directions API, Distance Matrix API, Elevation API, Geocoding API, and so on. See https://developers.google.com/maps/pricing-and-plans/ https://developers.google.com/maps/pricing-and-plans/
    - ching_wow_ka 11y ago
      I think you'll agree that there's a big jump from being able to access an API X times a day to having access to a data dump.
      - kaeawc 11y ago
        There are specific clauses in Google's APIs that allow for things like disaster recovery applications to have unlimited access for free.
        feld 11y ago
        So I can download their entire map dataset and host my own copy?
  - baby 11y ago
    > They are improving your life in exchange for information about you and the places you visit So they are improving the world and for free. I'm glad we all agree
    - glitchdout 11y ago
      information about you and the places you visit != free
      - superuser2 11y ago
        Not exactly free, but the places you visit can be observed with the naked eye from public property, and a network of cameras feeding into ANPR and facial recognition would be a hell of a lot cheaper (and more reliable) than a mapping product. Not to mention that realtime traffic data and intelligent routing around congestion is only workable by tracking individual vehicles on the road and recording how long it takes them to get from A to B.
- ekianjo 11y ago
  It would be better to contribute to OpenStreetMaps instead of force-feeding their own proprietary platform.
- visarga 11y ago
  In my country (Romania) it's hit or miss. Not all addresses can be located by Google Maps. Many of them yield just the street, which might be kilometers long. It's a shame that Google Maps is inferior to my iGo tablet running Windows Mobile from 10 years ago. And it didn't need to have continuous wireless data access. Even routes that have been looked up are not saved, so there is no way to recover the route if the app crashes or I need to interrupt it midway and I have no data signal. I am very frustrated with Google Maps as a car GPS. It's an inferior product in my country. How was iGo capable to deliver superior maps so many years ago?
  - r3bl 11y ago
    A bit further on the west (in Bosnia & Herzegovina) Google's street view is still not available (not even in our capital). As for the names of the streets, OpenStreetMap is far more detailed (and precise). Even Bing's Maps are more usable.
    - tanzam75 11y ago
      > It's a shame that Google Maps is inferior to my iGo tablet running Windows Mobile from 10 years ago. And it didn't need to have continuous wireless data access. ... How was iGo capable to deliver superior maps so many years ago? They purchased data from a mapping company. Ten years ago, the two largest mapping companies in the world were Navteq and Tele Atlas. Google used to buy data from these companies as well, but they stopped paying and decided to build up their own maps data from scratch. > Even Bing's Maps are more usable. If you zoom to Sarajevo and look at the bottom-right of the Bing Maps window, it says: "© 2015 HERE. © 2015 Microsoft Corporation." Navteq was bought by Nokia, got renamed HERE Maps, and has just been sold to a consortium of German automakers. Microsoft currently has a multi-year contract with HERE Maps as part of their deal with Nokia. Also, Helmut Panke is on the Microsoft Board of Directors. He used to be the Chairman of the Management Board of BMW.
    - discardorama 11y ago
      > A bit further on the west (in Bosnia & Herzegovina) Google's street view is still not available Could it be because of the maze of laws that governs public photography? In the US, it's somewhat simple: what's visible from the street can be photographed. But in other countries, you have murky laws, and Google, with deep pockets, can get sued easily.
- kuschku 11y ago
  In my region, their data is from 2005, with no MapMaker available, so people always end up in my suburb and ask where the freeway is. StreetView is only available in some areas, after Google put their cameras at a height of 3.5m, far above eye level, and needing written permission from everyone (cause people built hedges and fences with eyelevel in mind, so people might be in their frontyard, behind a hedge, naked, and now Google photographs that – big no-go). Just another thing where Google thought violating reasonable laws would be possible for them. In the end, their quality is hit and miss, and the cost is too high. No public transit data, not the ability to download larger areas, etc.
- lmm 11y ago
  If I worked on making the big mac tastier would you call that improving the world? I mean, plenty of people eat them and improving the flavour improves their lives. Ultimately the specific terminology is unimportant. Do we agree that there's a distinction between digitizing books in a way that's accessible to everyone, and working to improve Google products that exist primarily to increase Google's revenue, and can you see why people would want to contribute to one and not the other?
  - arihant 11y ago
    First, reCAPTCHA's purpose was to digitize books, not digitize them for public domain. If the work was copyrighted, it was returned to rightful owners. Second, if Big Mac got 2000% tastier every year, and while at that, 2 billion people got to eat dinner for free everyday, it sure will make the world better.
dzhiurgis 11y ago
Data mining books was the lowest hanging fruit. Being a library was a side effect. First and foremost was training their search engine with a massive amount of information.
- omonra 11y ago
  How do you train a search engine on books? How do you generate relevance metrics (like # of links on the web)?
  - Tycho 11y ago
    When it is able to offer search suggestions or corrections to your search, I think some of that was based on machine learning on a massive corpus of text.
    - omonra 11y ago
      I don't think so. I think it simply performs a spellcheck vs existing database of terms and offers a likely correction. Ie when you put in a word 'Mroscow' it sees that this term has 500 results while a similar term 'Moscow' has 100 million results. Therefore it assumes that you simply misspelled the word and offers the alternative as suggestion. It does not know what the actual underlying word means.
      - no_news_is 11y ago
        @Tycho http://norvig.com/spell-correct.html http://norvig.com/spell-correct.html
  - dzhiurgis 11y ago
    Perhaps same way people associate facts? PageRank is one of the 200 metrics Google uses nowadays. It's weight is not as great before.
arihant 11y ago
The world is not a place where events can be sandboxed to one instance like this. Luis von Ahn did digitize the entire archive of NYT before flipping reCAPTCHA to Google, and I'm sure it's still helping. But because he did that, he had time to start monolingo project (now duolingo) which is helping translate the Web. So if reCAPTCHA was not being turned into a tool for profit, it would be sitting in dust at some old room in Wean Hall. Or, Luis would be still with it, and Duolingo would never have existed. Besides, all work that Luis did on CAPTCHA at Berkeley and reCAPTCHA at CMU is published. Nobody is being prevented from making another one just for books. Amazon, maybe? reCAPTCHA is being utilized for other things because digitizing books in that manner is not useful anymore. Most publishers now have ebook versions of most of their books, which are already digitized.
- mbrutsch 11y ago
  > because digitizing books in that manner is not useful anymore. Most publishers now have ebook versions of most of their books, which are already digitized. Perhaps for kids looking for new books that's wonderful, but those of us that use books older than your mom, there are no ebooks. Google Books was literally the only place to access that information outside of a few select world-class university libraries.
  - arihant 11y ago
    I was referring to digitizing books as Luis defines it - actually typing in the scanned pages. Not the scanned pages themselves, as Google Books had. My comment was regarding reCAPTCHA not being used to convert scanned book pages to text. For books older than my mom, most of which must be public domain, there are a lot online archives. In fact, some of the more common archives like Project Gutenberg, are available as free ebooks on Kindle and iBooks. For more obscure titles, there are field-specific archives from where Google accessed them. They only got information from publishers and archives. So information you're looking for is out there somewhere in digital form.
    - mbrutsch 11y ago
      > So information you're looking for is out there somewhere in digital form. Were that true, you'd think Google would be able to find it.
visarga 11y ago
> identify the images with salads I tried many times and failed the salad test. It's so frustrating that I gave up.
- DanBC 11y ago
  "Identify the images with spinach" - presents 8 blurry green images of green spinach like leaves.
- rwmj 11y ago
  I failed the "salad test" too, I think because I identified various plates of mixed meat and non-green veg as salads. Probably too much time spent going to hip London restaurants.
- r3bl 11y ago
  IIRC the image recognition doesn't even show up if you just move your mouse randomly across the frame. I have been requested to identify pictures like two or three times in total.
baby 11y ago
> to improve Googles online mapping then > nothing to do with improving the world seems contradictory to me.
- deelowe 11y ago
  Seriously... How this the top comment? It's full of hyperbole and straight up misrepresentation of facts.
cbhl 11y ago
reCAPTCHA switched away from digitizing books simply because it was no longer a good test for human vs robot -- the robots were better at it than humans. "[Google's research shows that] today’s Artificial Intelligence technology can solve even the most difficult variant of distorted text at 99.8% accuracy. Thus distorted text, on its own, is no longer a dependable test" http://googleonlinesecurity.blogspot.com/2014/12/are-you-robot-introducing-no-captcha.html http://googleonlinesecurity.blogspot.com/2014/12/are-you-rob... But lots of sites depend on reCAPTCHA, it would have been unreasonable to just shut it down and break a huge swath of the internet because it had achieved the "we could digitize just about any book written in latin script" goal.
deelowe 11y ago
Do you have anything to back this up? - Google stopped using book scans for reCAPTCHA, because it stopped preventing spam - How can you claim Google's maps product hasn't improved the world? That's a bold statement. I have a data point: It was used quite a bit during katrina and fukashima. - The training data that's now being used for ML is helping the world of image and audio recognition immensely. They do publish a lot of this stuff, ya know: https://research.google.com/ https://research.google.com/

by chippy 11y ago

blfr 11y ago

How is reliable Google Maps and Street View not improving the world? I'm using them daily and they're improving my life a lot.

tdkl 11y ago
At what cost?
- golergka 11y ago
  At cost of solving the captcha problem?
blub 11y ago
They are improving your life in exchange for information about you and the places you visit. Their product is closed, the proprietary data is not accessible outside their apps, or if they wish to license it. OpenStreetMap is improving the world.
- mrb 11y ago
  "the proprietary data is not accessible outside their apps" This is not true. Many APIs are available, specifically to allow building 3rd-party apps: Directions API, Distance Matrix API, Elevation API, Geocoding API, and so on. See https://developers.google.com/maps/pricing-and-plans/ https://developers.google.com/maps/pricing-and-plans/
  - ching_wow_ka 11y ago
    I think you'll agree that there's a big jump from being able to access an API X times a day to having access to a data dump.
    - kaeawc 11y ago
      There are specific clauses in Google's APIs that allow for things like disaster recovery applications to have unlimited access for free.
      - feld 11y ago
        So I can download their entire map dataset and host my own copy?
- baby 11y ago
  > They are improving your life in exchange for information about you and the places you visit So they are improving the world and for free. I'm glad we all agree
  - glitchdout 11y ago
    information about you and the places you visit != free
    - superuser2 11y ago
      Not exactly free, but the places you visit can be observed with the naked eye from public property, and a network of cameras feeding into ANPR and facial recognition would be a hell of a lot cheaper (and more reliable) than a mapping product. Not to mention that realtime traffic data and intelligent routing around congestion is only workable by tracking individual vehicles on the road and recording how long it takes them to get from A to B.
ekianjo 11y ago
It would be better to contribute to OpenStreetMaps instead of force-feeding their own proprietary platform.
visarga 11y ago
In my country (Romania) it's hit or miss. Not all addresses can be located by Google Maps. Many of them yield just the street, which might be kilometers long. It's a shame that Google Maps is inferior to my iGo tablet running Windows Mobile from 10 years ago. And it didn't need to have continuous wireless data access. Even routes that have been looked up are not saved, so there is no way to recover the route if the app crashes or I need to interrupt it midway and I have no data signal. I am very frustrated with Google Maps as a car GPS. It's an inferior product in my country. How was iGo capable to deliver superior maps so many years ago?
- r3bl 11y ago
  A bit further on the west (in Bosnia & Herzegovina) Google's street view is still not available (not even in our capital). As for the names of the streets, OpenStreetMap is far more detailed (and precise). Even Bing's Maps are more usable.
  - tanzam75 11y ago
    > It's a shame that Google Maps is inferior to my iGo tablet running Windows Mobile from 10 years ago. And it didn't need to have continuous wireless data access. ... How was iGo capable to deliver superior maps so many years ago? They purchased data from a mapping company. Ten years ago, the two largest mapping companies in the world were Navteq and Tele Atlas. Google used to buy data from these companies as well, but they stopped paying and decided to build up their own maps data from scratch. > Even Bing's Maps are more usable. If you zoom to Sarajevo and look at the bottom-right of the Bing Maps window, it says: "© 2015 HERE. © 2015 Microsoft Corporation." Navteq was bought by Nokia, got renamed HERE Maps, and has just been sold to a consortium of German automakers. Microsoft currently has a multi-year contract with HERE Maps as part of their deal with Nokia. Also, Helmut Panke is on the Microsoft Board of Directors. He used to be the Chairman of the Management Board of BMW.
  - discardorama 11y ago
    > A bit further on the west (in Bosnia & Herzegovina) Google's street view is still not available Could it be because of the maze of laws that governs public photography? In the US, it's somewhat simple: what's visible from the street can be photographed. But in other countries, you have murky laws, and Google, with deep pockets, can get sued easily.
kuschku 11y ago
In my region, their data is from 2005, with no MapMaker available, so people always end up in my suburb and ask where the freeway is. StreetView is only available in some areas, after Google put their cameras at a height of 3.5m, far above eye level, and needing written permission from everyone (cause people built hedges and fences with eyelevel in mind, so people might be in their frontyard, behind a hedge, naked, and now Google photographs that – big no-go). Just another thing where Google thought violating reasonable laws would be possible for them. In the end, their quality is hit and miss, and the cost is too high. No public transit data, not the ability to download larger areas, etc.
lmm 11y ago
If I worked on making the big mac tastier would you call that improving the world? I mean, plenty of people eat them and improving the flavour improves their lives. Ultimately the specific terminology is unimportant. Do we agree that there's a distinction between digitizing books in a way that's accessible to everyone, and working to improve Google products that exist primarily to increase Google's revenue, and can you see why people would want to contribute to one and not the other?
- arihant 11y ago
  First, reCAPTCHA's purpose was to digitize books, not digitize them for public domain. If the work was copyrighted, it was returned to rightful owners. Second, if Big Mac got 2000% tastier every year, and while at that, 2 billion people got to eat dinner for free everyday, it sure will make the world better.

dzhiurgis 11y ago

Data mining books was the lowest hanging fruit. Being a library was a side effect. First and foremost was training their search engine with a massive amount of information.

omonra 11y ago
How do you train a search engine on books? How do you generate relevance metrics (like # of links on the web)?
- Tycho 11y ago
  When it is able to offer search suggestions or corrections to your search, I think some of that was based on machine learning on a massive corpus of text.
  - omonra 11y ago
    I don't think so. I think it simply performs a spellcheck vs existing database of terms and offers a likely correction. Ie when you put in a word 'Mroscow' it sees that this term has 500 results while a similar term 'Moscow' has 100 million results. Therefore it assumes that you simply misspelled the word and offers the alternative as suggestion. It does not know what the actual underlying word means.
    - no_news_is 11y ago
      @Tycho http://norvig.com/spell-correct.html http://norvig.com/spell-correct.html
- dzhiurgis 11y ago
  Perhaps same way people associate facts? PageRank is one of the 200 metrics Google uses nowadays. It's weight is not as great before.

arihant 11y ago

The world is not a place where events can be sandboxed to one instance like this. Luis von Ahn did digitize the entire archive of NYT before flipping reCAPTCHA to Google, and I'm sure it's still helping. But because he did that, he had time to start monolingo project (now duolingo) which is helping translate the Web. So if reCAPTCHA was not being turned into a tool for profit, it would be sitting in dust at some old room in Wean Hall. Or, Luis would be still with it, and Duolingo would never have existed. Besides, all work that Luis did on CAPTCHA at Berkeley and reCAPTCHA at CMU is published. Nobody is being prevented from making another one just for books. Amazon, maybe? reCAPTCHA is being utilized for other things because digitizing books in that manner is not useful anymore. Most publishers now have ebook versions of most of their books, which are already digitized.

mbrutsch 11y ago
> because digitizing books in that manner is not useful anymore. Most publishers now have ebook versions of most of their books, which are already digitized. Perhaps for kids looking for new books that's wonderful, but those of us that use books older than your mom, there are no ebooks. Google Books was literally the only place to access that information outside of a few select world-class university libraries.
- arihant 11y ago
  I was referring to digitizing books as Luis defines it - actually typing in the scanned pages. Not the scanned pages themselves, as Google Books had. My comment was regarding reCAPTCHA not being used to convert scanned book pages to text. For books older than my mom, most of which must be public domain, there are a lot online archives. In fact, some of the more common archives like Project Gutenberg, are available as free ebooks on Kindle and iBooks. For more obscure titles, there are field-specific archives from where Google accessed them. They only got information from publishers and archives. So information you're looking for is out there somewhere in digital form.
  - mbrutsch 11y ago
    > So information you're looking for is out there somewhere in digital form. Were that true, you'd think Google would be able to find it.

visarga 11y ago

> identify the images with salads I tried many times and failed the salad test. It's so frustrating that I gave up.

DanBC 11y ago
"Identify the images with spinach" - presents 8 blurry green images of green spinach like leaves.
rwmj 11y ago
I failed the "salad test" too, I think because I identified various plates of mixed meat and non-green veg as salads. Probably too much time spent going to hip London restaurants.
r3bl 11y ago
IIRC the image recognition doesn't even show up if you just move your mouse randomly across the frame. I have been requested to identify pictures like two or three times in total.

baby 11y ago

> to improve Googles online mapping then > nothing to do with improving the world seems contradictory to me.

deelowe 11y ago
Seriously... How this the top comment? It's full of hyperbole and straight up misrepresentation of facts.

cbhl 11y ago

reCAPTCHA switched away from digitizing books simply because it was no longer a good test for human vs robot -- the robots were better at it than humans. "[Google's research shows that] today’s Artificial Intelligence technology can solve even the most difficult variant of distorted text at 99.8% accuracy. Thus distorted text, on its own, is no longer a dependable test" http://googleonlinesecurity.blogspot.com/2014/12/are-you-robot-introducing-no-captcha.html http://googleonlinesecurity.blogspot.com/2014/12/are-you-rob... But lots of sites depend on reCAPTCHA, it would have been unreasonable to just shut it down and break a huge swath of the internet because it had achieved the "we could digitize just about any book written in latin script" goal.

deelowe 11y ago

Do you have anything to back this up? - Google stopped using book scans for reCAPTCHA, because it stopped preventing spam - How can you claim Google's maps product hasn't improved the world? That's a bold statement. I have a data point: It was used quite a bit during katrina and fukashima. - The training data that's now being used for ML is helping the world of image and audio recognition immensely. They do publish a lot of this stuff, ya know: https://research.google.com/ https://research.google.com/