3 ms·
> the system knows "Tylenol" is synonymous with "Paracetamol", "Acetaminophen", etc. This is the exactly the kind of thing that Google cannot fathom manually d
by titzer 5y ago
> the system knows "Tylenol" is synonymous with "Paracetamol", "Acetaminophen", etc.
This is the exactly the kind of thing that Google cannot fathom manually doing. As if entering facts into a computer were morally wrong somehow. They'd much rather launch the equivalent of a shell script that harnesses face-melting amounts of computational power, processing literally trillions of webpages in bulk, signal and noise together, junk, spam, and misdirection alike, to learn bad associations and then serve them up with no human review and then put the full force of their reputation behind a results page that apparently people never check and certainly can't correct because of the inscrutability of a machine-learned model that has few to no levers to adjust.
- PaulHoule 5y agoGoogle has long lied about what they do. I had a chance to debrief people who had left their relevance team and they told me things that were outright contradictory to what rank-and-file Google employees have told me. (What they told me did make sense in terms of my experience as an IR system developer, SEO publisher, etc.) Microsoft bought a company called PowerSet that had extracted a large database of entities and relationships from Wikipedia and used the technology to make the "Bing" search engine. Earlier Microsoft engines were a joke, but Bing was so good that Google saw it as a threat so they bought Freebase to get a similar kind of database, then they killed it to incorporate it into the "Google Knowledge Graph". For all of their hating on semantics note that they hired R. V. Guha as their chief scientist, who worked with Doug Lenat on the notorious https://en.wikipedia.org/wiki/Cyc https://en.wikipedia.org/wiki/Cyc
- leobg 5y agoIsn’t that kind of easy to do? I mean, you can do that kind of thing as a one-man-show on consumer hardware using GloVe or fastText.