3 ms·

It's most valuable feature is the novelness of it - it is likely being used by over a million people right now. I assume that Apple is tracking result click-th

It's most valuable feature is the novelness of it - it is likely being used by over a million people right now. I assume that Apple is tracking result click-throughs to guess when Siri is providing (or actually, not providing) the right answer to your query. This should mean it'll become exponentially better over time.

ramchip 15y ago
I'd expect it to be more like a logarithmic increase than an exponential one. The asymptote is the point where you reach perfect integration or fine tuning of the algorithm, at which point extra feedback stops being useful. Or perhaps I'm just taking "exponentially" too literally...
- CamperBob 15y ago
  How does exponential growth differ from logarithmic growth? An exponent is a logarithm, no...?
  - xxbondsxx 15y ago
    Woof! Someone's been out of high school for too long :P exp(x) takes e (the constant) and raises it to the x power. Log(x) finds the power to raise the base (usually 10 but sometimes e) to produce x. They are essentially inverses of each other. Log(exp(x)) = x; Graphically they are quite different: http://www.wolframalpha.com/input/?i=exp%28x%29 http://www.wolframalpha.com/input/?i=exp%28x%29 http://www.wolframalpha.com/input/?i=ln%28x%29 http://www.wolframalpha.com/input/?i=ln%28x%29
    - woof 15y ago
      You called?
    - CamperBob 15y ago
      Sure, but that's just a matter of how the terminology is used. Setting aside natural logs for a moment, consider that log10(1000) is 3, where 3 is simultaneously the "exponent" to which 10 must be raised to equal 1000, and the "logarithm," in the sense that it's the result of the log10 function, as might be found by looking in a table of logarithms.
      - ramchip 15y ago
        I understand what you mean, but the problem is that the terminology has a precisely defined meaning, and if you ignore the convention you will be misunderstood. It would be like saying that a certain O(log n) algorithm has an exponential complexity -- the meaning is completely different. The growth of a function concerns the dependent variable, so y=e^x has exponential growth, while y=ln(x) is logarithmic. See also: http://en.wikipedia.org/wiki/Logarithmic_growth http://en.wikipedia.org/wiki/Logarithmic_growth http://en.wikipedia.org/wiki/Exponential_growth http://en.wikipedia.org/wiki/Exponential_growth (Sorry for the off-topic tangent!)

ramchip 15y ago

I'd expect it to be more like a logarithmic increase than an exponential one. The asymptote is the point where you reach perfect integration or fine tuning of the algorithm, at which point extra feedback stops being useful. Or perhaps I'm just taking "exponentially" too literally...

CamperBob 15y ago
How does exponential growth differ from logarithmic growth? An exponent is a logarithm, no...?
- xxbondsxx 15y ago
  Woof! Someone's been out of high school for too long :P exp(x) takes e (the constant) and raises it to the x power. Log(x) finds the power to raise the base (usually 10 but sometimes e) to produce x. They are essentially inverses of each other. Log(exp(x)) = x; Graphically they are quite different: http://www.wolframalpha.com/input/?i=exp%28x%29 http://www.wolframalpha.com/input/?i=exp%28x%29 http://www.wolframalpha.com/input/?i=ln%28x%29 http://www.wolframalpha.com/input/?i=ln%28x%29
  - woof 15y ago
    You called?
  - CamperBob 15y ago
    Sure, but that's just a matter of how the terminology is used. Setting aside natural logs for a moment, consider that log10(1000) is 3, where 3 is simultaneously the "exponent" to which 10 must be raised to equal 1000, and the "logarithm," in the sense that it's the result of the log10 function, as might be found by looking in a table of logarithms.
    - ramchip 15y ago
      I understand what you mean, but the problem is that the terminology has a precisely defined meaning, and if you ignore the convention you will be misunderstood. It would be like saying that a certain O(log n) algorithm has an exponential complexity -- the meaning is completely different. The growth of a function concerns the dependent variable, so y=e^x has exponential growth, while y=ln(x) is logarithmic. See also: http://en.wikipedia.org/wiki/Logarithmic_growth http://en.wikipedia.org/wiki/Logarithmic_growth http://en.wikipedia.org/wiki/Exponential_growth http://en.wikipedia.org/wiki/Exponential_growth (Sorry for the off-topic tangent!)