4 ms·

Thirdly, Mr Cook should put some of that $250b cash mountain he sits on to work making this stuff happen. Prove that localized models built on small datasets w

Thirdly, Mr Cook should put some of that $250b cash mountain he sits on to work making this stuff happen. Prove that localized models built on small datasets work. Invest in cryptography and user anonymity technologies. Instead he's just sitting on it like smaug.

saagarjha 8y ago
> Prove that localized models built on small datasets work. > Invest in cryptography and user anonymity technologies. That’s what they literally do. Much of iOS’s “proactive” features are done on-device, as is things like photos analysis. And most of Apple’s products utilize at least some sort of cryptography and anonymized logging.
- huffmsa 8y ago
  Clearly they're not doing enough, if they were, Siri wouldn't have fallen so far behind. I know it's a new and difficult field (I work in it) but it's also one where you could throw buckets of cash at the problem and get better results.
  - theshrike79 8y ago
    Siri is far behind because Apple has a pro-privacy stance. They've effectively hamstrung themselves by anonymising the data and not cross-correlating between users as much as Google does. Google on the other hand has no qualms about using the mountains of data they collect on every single user on their platforms. They can teach Google Assistant stuff that Siri can't even try to do yet.
    - kkarakk 8y ago
      siri is behind because apple thought buying a natural language assistant startup was enough work/money spent to get into the smart assistant game. there is literally no research coming out of apple on smart assistants or machine learning(that isn't photos) and it shows
      - saagarjha 8y ago
        > there is literally no research coming out of apple on smart assistants or machine learning(that isn't photos) https://machinelearning.apple.com/ https://machinelearning.apple.com/

4 ms·

Thirdly, Mr Cook should put some of that $250b cash mountain he sits on to work making this stuff happen. Prove that localized models built on small datasets w

by huffmsa 8y ago

saagarjha 8y ago
> Prove that localized models built on small datasets work. > Invest in cryptography and user anonymity technologies. That’s what they literally do. Much of iOS’s “proactive” features are done on-device, as is things like photos analysis. And most of Apple’s products utilize at least some sort of cryptography and anonymized logging.
- huffmsa 8y ago
  Clearly they're not doing enough, if they were, Siri wouldn't have fallen so far behind. I know it's a new and difficult field (I work in it) but it's also one where you could throw buckets of cash at the problem and get better results.
  - theshrike79 8y ago
    Siri is far behind because Apple has a pro-privacy stance. They've effectively hamstrung themselves by anonymising the data and not cross-correlating between users as much as Google does. Google on the other hand has no qualms about using the mountains of data they collect on every single user on their platforms. They can teach Google Assistant stuff that Siri can't even try to do yet.
    - kkarakk 8y ago
      siri is behind because apple thought buying a natural language assistant startup was enough work/money spent to get into the smart assistant game. there is literally no research coming out of apple on smart assistants or machine learning(that isn't photos) and it shows
      - saagarjha 8y ago
        > there is literally no research coming out of apple on smart assistants or machine learning(that isn't photos) https://machinelearning.apple.com/ https://machinelearning.apple.com/