4 ms·

Click tracking is MUCH more significant than ML/AI, Google results were very good even 10 years back when they were just based on click tracking. Whatever they

Click tracking is MUCH more significant than ML/AI, Google results were very good even 10 years back when they were just based on click tracking. Whatever they are doing to parse/steal other websites data to show as snippets is only complicated because there is no schema. Look what Amazon did, they made people implement "skill" so it's all API now. All that was solved for majority of use cases and leaped ahead by Amazon in under 5 years with Alexa, deployed as tiny hardware. Remove click tracking and top 1000 sites from Google and it will be shit, no amount of ML/AI dance will solve that. Apple can just do top 1000 sites from each country with click tracking (to track which result users are preferring) and significantly cut traffic sent to Google. They already try to do this with Siri. Bing was chicken(user) and egg problem (click tracking to optimize results).

crazygringo 6y ago
I'm sorry, but nothing of what you're saying makes any sense to me. All click tracking does is measure the quality of your results -- whether a user clicks on the first result or third or none at all. It doesn't help generate quality results in the first place. Google's main first innovation was PageRank, not click tracking. Click tracking is just a metric to judge the underlying algorithms. (There is one exception: it can be used to fine-tune the order of the top 5-10 results for the most popular queries, but of course most of search is in the long tail where that fine tuning can't be done. And in any case, you still need excellent candidates for your first 10 items to begin with.) Click tracking does nothing for parsing a query, nothing for ranking the top 10 items out of a million matches, and nothing for extracting information. And why are you talking about the top 1,000 sites? My whole point is crawling the web is the easy part.

crazygringo 6y ago

I'm sorry, but nothing of what you're saying makes any sense to me. All click tracking does is measure the quality of your results -- whether a user clicks on the first result or third or none at all. It doesn't help generate quality results in the first place. Google's main first innovation was PageRank, not click tracking. Click tracking is just a metric to judge the underlying algorithms. (There is one exception: it can be used to fine-tune the order of the top 5-10 results for the most popular queries, but of course most of search is in the long tail where that fine tuning can't be done. And in any case, you still need excellent candidates for your first 10 items to begin with.) Click tracking does nothing for parsing a query, nothing for ranking the top 10 items out of a million matches, and nothing for extracting information. And why are you talking about the top 1,000 sites? My whole point is crawling the web is the easy part.