5 ms·

I admire Google, but as a China-based technologist with years in Beijing it has been clear to everyone I know for a long time that Google China was an organizat

I admire Google, but as a China-based technologist with years in Beijing it has been clear to everyone I know for a long time that Google China was an organizational and strategic disaster. Levy paints this as a political problem. But really it is an organizational one. The decision to hire Kai Fu Lee was a lurching move made impetuously from the top. And when they pulled out it was a lurching move made impetuously from the top. The biggest challenge for Google China was never improving Google's core algorithms and services (that stuff was sensibly walled off). It was justifying long-term involvement in China. But Google China didn't innovate. It didn't compete. And it failed to produce or champion a single project interesting or "non-evil" enough to persuade its disconnected Mountain View overlords that a public dust-up with the CCP might be a bad thing.

vorg 16y ago
Google's IME (input method editor) for Chinese characters was the first of its kind. It's been cloned by Chinese businesses, and it had its initial teething problems re the data problem, but I still use it when on the web from Windows.
- trevelyan 16y ago
  Perfect example. The Google IME was nowhere near the first of its kind; it just became popular with non-native speakers because Google didn't cram in advertising like its Chinese predecessors and made it free. The algorithms involved are technically simple and the basic data structures are multigrams indexed as a binary tree or HMMs. The fact that developers had to steal data to do it instead of rolling their own corpus is a perfect example of the sort of chronic failure I'm talking about. Another is the fact they licensed Kingsoft materials (built off Linguistic Data Consortium data) instead of developing their own. Why outsource data development when you can hire people for under $2k per month to just do it for you? The English language media typically focuses on Google's genuine political problems because it wants to use the company as an indicator for the state of US-China relations. So people slam Baidu (an innovative company!) as a copycat and no-one remarks on the way Google's behavior suggests no serious effort or willingness to tackle the non-trivial or data-intensive problems that crop up in the semantic analysis of Chinese text. It is possible Google was doing really innovative stuff that I haven't heard about, but the fact that Mountain View made the decision to pull out of the country without even informing its China office is yet another data point suggesting the China office wasn't doing anything terribly worthwhile. Google shut house the way you'd fire an unproductive staff member. At a minimum, this lack of execution and/or attention to competing through innovation is embarrassing. Not for the main office which is focusing on language agnostic technologies. But it is bad for Google China and I would even go so far as to call it borderline incompetent. You can bet the parent company is heavily invested in semantic search in English, and Chinese is arguably a more important language for text analysis and retrieval in the 21st century.

vorg 16y ago

Google's IME (input method editor) for Chinese characters was the first of its kind. It's been cloned by Chinese businesses, and it had its initial teething problems re the data problem, but I still use it when on the web from Windows.

trevelyan 16y ago
Perfect example. The Google IME was nowhere near the first of its kind; it just became popular with non-native speakers because Google didn't cram in advertising like its Chinese predecessors and made it free. The algorithms involved are technically simple and the basic data structures are multigrams indexed as a binary tree or HMMs. The fact that developers had to steal data to do it instead of rolling their own corpus is a perfect example of the sort of chronic failure I'm talking about. Another is the fact they licensed Kingsoft materials (built off Linguistic Data Consortium data) instead of developing their own. Why outsource data development when you can hire people for under $2k per month to just do it for you? The English language media typically focuses on Google's genuine political problems because it wants to use the company as an indicator for the state of US-China relations. So people slam Baidu (an innovative company!) as a copycat and no-one remarks on the way Google's behavior suggests no serious effort or willingness to tackle the non-trivial or data-intensive problems that crop up in the semantic analysis of Chinese text. It is possible Google was doing really innovative stuff that I haven't heard about, but the fact that Mountain View made the decision to pull out of the country without even informing its China office is yet another data point suggesting the China office wasn't doing anything terribly worthwhile. Google shut house the way you'd fire an unproductive staff member. At a minimum, this lack of execution and/or attention to competing through innovation is embarrassing. Not for the main office which is focusing on language agnostic technologies. But it is bad for Google China and I would even go so far as to call it borderline incompetent. You can bet the parent company is heavily invested in semantic search in English, and Chinese is arguably a more important language for text analysis and retrieval in the 21st century.