[FoRK] Stand up and face the future
jbone at place.org
Wed Dec 23 04:19:54 PST 2009
One other bit:
> Notice I don't really like to use the term "realtime" because it
> never is.
I feel ya on this one. However, at the edges of this stuff --- even
in non-physical apps like the kind I deal with, much less robotic and
vehicular control apps and so on --- "real time" (at least to the
extent of being able to determine "when" some thing happened relative
to some other thing from some specific point of view, i.e. meaningful
time stamp generation) is now measured at precision *below that of the
clock cycle of the hardware it's on* and reaction times become
frequently critical in the microsecond range, with nano looming. And
for those "when did X happen" problems, and with a close eye to
special relativity in the multi-machine case, even GPS-enabled NTP has
become too "sloppy" and you find things like PTP and other, more
precise protocols cropping up. (Even that is ultimately a losing
battle because, in the limit, the notions of simultaneity and global
total ordering are non-physically meaningful.)
IME, RT is --- and for some time has been --- more about whether the
hardware / software stack is such that it permits earliest-possible
reaction to events --- i.e., user tasks can pre-empt system tasks when
necessary --- rather than the slippery slope of precision.
On Dec 23, 2009, at 6:10 AM, Jeff Bone wrote:
> Ken says:
>> So my $0.02 is that instead of changing the definition of what a
>> mine is, just drop it and talk about sifting.
> Often, the math is the same in "mining" and "sifting" --- it's just
> a matter of how you feed and store the data. (Often the math
> *isn't* the same, but that's a different story.)
> So no, I'm not particularly a "semantics" glosser; but in this
> case, there's inadequate precision in the terminology already, to
> the point that the distinction is useless...
> On Dec 22, 2009, at 9:44 PM, Jeff Bone wrote:
>> Ken says:
>>> But you can't "mine" in realtime. Mining requires a pre-existing
>>> source of accumulated data. Or it's not mining. By definition.
>> Ah, glasshoppah... ;-)
>> Cf. "online learning." As opposed to "supervised..."
>> My caveat (already stipulated) is that while it's true that today,
>> for the most part, the two are clearly separable activities, this
>> is already becoming a blurry distinction about the edges --- and
>> will become even moreso.
>> It's true that the notion of "mining" is useless without some
>> historical corpus to work against, but the distinction between that
>> historical corpus and the stream of things coming in in real time
>> is increasingly meaningless. I.e., you can pick any definitions
>> you want, but I can tell you from recent and relevant experience
>> that the definitions you are asserting above are arbitrary and not
>> very durable.
>> On Dec 22, 2009, at 8:21 AM, Jeff Bone wrote:
>>> Stephen says:
>>>> Mining, sifting, and real-time do not go together. At least not
>>>> in the traditional forms.
>>> With the caveat that mining and sifting / reacting in realtime are
>>> (today) different activities, I have to point at the existence
>>> proof (what I do for a living these days) as contrary to this
>>> point of view. And that activity and its general form are going
>>> to become even more prevalent outside the domain I operate in as
>>> we realize an Internet of Things. (It's already the case that
>>> this happens elsewhere; in the UK to a large extent and in the US
>>> to a much lesser extent realtime face recognition over public
>>> camera networks occurs already today...)
More information about the FoRK