3 ms·

I totally agree with this. Some people may say we need to cover every possible situation in a system, but this is true in theory, but practically we need some t

I totally agree with this. Some people may say we need to cover every possible situation in a system, but this is true in theory, but practically we need some trade offs when designing a system. We have to calculate the gains and costs while considering fixing the complexities.

archgoon 11y ago
The catch tends to be if you're able to detect when you're in a more complicated situation and bail out accordingly. One very useful thing is auto-labeling, which turns out, getting it wrong is more problematic than one would expect. Take for example the 'Gorilla' issue that Google encountered. http://nyti.ms/1JlTBab http://nyti.ms/1JlTBab [www.nytimes.com] Then there was the issue with the bookmarks and suicide categories http://bit.ly/1JFkmCr http://bit.ly/1JFkmCr [productforums.google.com] (This one isn't as big of a problem, as the frustration was probably more due to the compulsory usage of the new bookmarks, but it still shows some hazards of auto-categorization). With social things, it can sometimes be very hard to solve a problem that works for 95% of the population, but doesn't piss off the remaining 5% (often with good justification). Unfortunately, detecting when you're in that situation in the first place is sometimes equivalent to the problem you're trying to solve.
- github-cat 11y ago
  I don't understand why a big company like Google can put some feature online if it's not thoroughly tested or thought. It is because it is a web product and it wants the users to be the testers?
  - archgoon 11y ago
    What would the testing plan look like?
    - icebraining 11y ago
      Simple procedure: take a large sample (1M pictures or so), tag them with the system and then show each picture with the label to a person (using Mechanical Turk or similar) and ask "is this label offensive?" Use simple statistics to identify trouble spots.
- sp332 11y ago
  Please don't use link shorteners. They just obfuscate the link target.

archgoon 11y ago

The catch tends to be if you're able to detect when you're in a more complicated situation and bail out accordingly. One very useful thing is auto-labeling, which turns out, getting it wrong is more problematic than one would expect. Take for example the 'Gorilla' issue that Google encountered. http://nyti.ms/1JlTBab http://nyti.ms/1JlTBab [www.nytimes.com] Then there was the issue with the bookmarks and suicide categories http://bit.ly/1JFkmCr http://bit.ly/1JFkmCr [productforums.google.com] (This one isn't as big of a problem, as the frustration was probably more due to the compulsory usage of the new bookmarks, but it still shows some hazards of auto-categorization). With social things, it can sometimes be very hard to solve a problem that works for 95% of the population, but doesn't piss off the remaining 5% (often with good justification). Unfortunately, detecting when you're in that situation in the first place is sometimes equivalent to the problem you're trying to solve.

github-cat 11y ago
I don't understand why a big company like Google can put some feature online if it's not thoroughly tested or thought. It is because it is a web product and it wants the users to be the testers?
- archgoon 11y ago
  What would the testing plan look like?
  - icebraining 11y ago
    Simple procedure: take a large sample (1M pictures or so), tag them with the system and then show each picture with the label to a person (using Mechanical Turk or similar) and ask "is this label offensive?" Use simple statistics to identify trouble spots.
sp332 11y ago
Please don't use link shorteners. They just obfuscate the link target.