Note that when I say "tends to be" I mean that, depending on the details of a social network, the scaling law can differ. In particular a network that relies on relative strangers having interactions with other relative strangers, as happens with eBay and Airbnb, is going to scale much closer to Metcalfe's Law. By contrast one that depends on developing a circle of friends, such as happens with Facebook, the scaling law will be closer to n log(n).
Disclaimer, I have a bias here since I'm one of the co-authors of the n log(n) law. (Andrew Odlyzko did all of the work. And then we found out that Bob Briscoe had independently arrived at the same conclusion based on data that he had access to at British Telecom.)
The trouble I have with Metcalfe's Law and similar is that all of them
are far too optimistic. Most English speakers have heard the phrase,
"Point of diminishing returns," and such points do exist. Worse yet,
there's also the phrase "Collapse under its own weight," describing the
point where something got so big or widespread to retain its integrity,
so its groth essentially harmed itself and failed.
If you look around, you'll see this kind of diminishing returns and self
destruction in a lot of different places, from "social" environments
like this forum, to your over-flowing email box, to the constant
interruption of your cell phone.
It is definitely true that a vocal minority of users are destructive to any network they find themselves in. Examples of negative users include things like trolls and spammers.
Another problem is that the natural desire of the people running the network to monetize it creates incentives to weaken the experience.
Both of these can lead to sub-linear growth in overall value as a network grows. But measuring value is already hard. Modeling all of the ways to screw up a network is even harder, and I doubt that there is any simple approximate general law that describes it.
Same here. I've never seen a simple general law, rule, or even guideline
to accurately predict the stagnation and decline of complex systems, or
more accurately, growth-decay cycles. It's a bit like knowing the sun
regularly rises somewhere over there (jcr points vaguely towards the
"East-ish" direction) but having no idea why it does, and hence,
having no way to accurately predict where it will rise. Trying to figure
it out can be both fascinating and frustrating.
There was a mathematician or scientist who said, "To measure something
is to know it," but unfortunately, I can't remember his name. Anyhow, I
agree our inability to accurately measure and model (or even notice) a
lot of the factors involved in complex systems results in our inability
to describe or predict them.
BTW, I kind of look at "measuring value" and "modeling ways to screw up
the network" to be mostly the same thing. In one case you're
identifying, measuring, and modeling the beneficial (value-increasing)
factors, and in other, you're identifying, measuring, and modeling the
harmful (value-decreasing) factors. --I have a funny feeling that I've
missed some thing obvious, so did I misunderstand your statement?
Our inability to accurately measure and model (or even notice) a lot of the factors involved in complex systems results in our inability to describe or predict them
Structured processes | SDIC [1] vs Resolution is a legitmate issue. Not at all processes can be modeled the same, like a "complicated" but ultimately simple one. In the latter, Logic helps "bridge" the resolution issues. Many deterministic processes cannot be brute-forced, though.
[1] ie, displaying Sensitive dependence upon initial conditions; just how accurate is your measure and can it ever be accurate enough to deduce an originating function?
Edit: If I may expand on the point above. Which may seem cryptic.
Complex systems are interisting in that they can be both predictable (in theory) and not predictable (in practice). What's more is that they can be mischaracterized by analyzing the data. Ie, this data is a mess, it must me unpredictable. There are a whole class of deterministic processes which generate these types of results. Whereas a normal, linear process can be infered from medium/high resolution data, even with higher-resolution data we can't infer the underlying logic of the complex system. We may, as a result, either oversimplify the model or proclaim the data to not support any deterministic process at all. The classic example is the class of processes that exhibit sensitive dependance upon initial conditions. ie, variations on the notion of deterministic Chaos. Their sesitivity is such that the resolution of the dataset required to deduce or infer the origination function would never be feasible if it was anything other than complete. Whereas, with deterministic processes that are more traditionally tractable, you can make progress in your knowledge with data-sets of increasing resolution. ie, you can run a regression to infer y=mx+b or a monte-carlo to fit a gaussian curve or what not. But you cannot "brute" force a fit to a choatic process from a montecarlo, becaue you will never have enough resolution nor enough precision in your data set to infer an origination function. [1]
The summary thought is tha sometimes gaps in data can be bridged with higher-level logic or heristic, but this is not always possible (either in theory or practice). Yet, we should not infer a problem is unsolvable or untractable just because of this. =D
Some complex systems can be described using simple laws e.g., complex interactions of gas molecules can be described using thermodynamics with aggregate terms such as temperature, pressure.
I am consistently amazed by the durability of this "cloud" hype. I mean, it's been at the peak of the hype cycle now for what, six years? I have been expecting a backlash for most of that time, and even now, when I make comments that are not glowing about "cloud" I get funny looks.
Part of this is that 'the cloud' is now how we refer to perfectly ordinary services that have been around as long as the web, like FTP space. I mean, I walked in on a group of guys discussing their 'time machine' backup scheme for their mac. Now, being a service provider with a whole lot of disk, my first thought, of course, was "can I sell some kind of standards-based something that will let people time-machine to my hard drives in a datacenter?" I got about half way through asking if such a thing existed, and they said, "Oh, you mean backup to the cloud?" at which an involuntary scoffing noise escaped my throat.
(the upshot is that apparently you can backup the time machine data to 'the cloud' but the people I talked to did not know what mechanism was used to upload/store the data.)
I think what we're observing is just the evolution of marketing to
people with increasingly lower amounts of technical acumen. At one
point, "on the system" was understood to mean something is stored on the
time-sharing mainframe, simply because everyone used dumb terminals to
access the infamous "system". As technology progressed and computers
became smaller, more affordable, and more capable, the lingo changed to
"on the server" since you actually had the capacity to store something
on your "system". The odd part is we still expected the "server" to be
located locally (in building, on campus, &c.), When a server wasn't
located (implied) "locally", it was called an "off site" server and was
typically accessed over private, dedicated lines/connections. This lingo
survived for a while when access methods changed to using connections
over the public Internet -- TYPICALLY DRAWN AS A CLOUD IN DIAGRAMS!
Some marketing person who didn't really understand technical diagrams of
networks saw the amorphous "cloud" representing the Internet in the
drawings and started referring to "the Internet" as "the cloud."
Needless to say, this lingo caught on with the non-technical folks, and
has been used ever since. I sincerely doubt we'll get rid of the phrase
any time soon since the majority of people on this planet are
non-technical and they tend to use the simple "hype" names they have
leaned.
The most interesting bit to all this linguistic history is you can
accurately profile people with it. For example, "Did he capitalize
'Internet'?" --definitely an old fart.
No, we've already lost to the people using "the cloud".
When you've got players like Microsoft, IBM, Apple, Oracle, Cisco, Amazon, Sales Force, etc - all referring to important parts of what they do as being "the cloud" (and all having different explanations of what "cloud computing" is), it's clear the the terminology is here to stay, no matter what it's origins or the intents/uses people once had for it.
If you read my post again, it seems we're "violently agreeing".
As I said, I doubt we'll get rid of the phrase.
Then again, from this point forward, I have every intention of
depicting the Internet in network drawings as a fishing net
labeled "Net" with the hope of tricking the marketing types into
using "In the Net" rather than "In the Cloud". I doubt it will
work, but it would still be fun to try. ;)
Heh. I think I might start drawing Visio diagrams with what used to be labelled "the cloud" instead labelled "the marketing department". When anybody asks, I'll say "nobody knows or can explain what happens in there or why, we just know that you put data in, and usually _something_ arrives at the destination. Company policy dictates that we have to use it, so we send all our data through it in ways where we can detect and correct any changes made in transit" ;-)
I do believe it is because it already worked before -- the cloud essentially needed one thing, cheap and fast consumer broadband.
3D printing needs a lot more than that, before it will become viable, while corporate blogging sounds a lot less useful than a twitting refrigriator (which could be a cool, if terribly expensive, gimick). After all, who the fuck wants to read a corporate blog from the people who make your toilet paper?
Changing definitions of the cloud? As far as I can tell the only definitions of the cloud that ever existed were basically:
1. On the Internet in a general sense. Somehow "cloud" became synonymous with Internet for many people and I can't figure out why. The weird part is that people who use "cloud" to mean Internet also use the term Internet to describe the same services the "cloud" offers. To me this makes me think marketing made this so. Somehow Dropbox is a cloud service but gmail (or hotmail or whatever email provider) is on the Internet according to many. How is that?
2. Definition 2 is basically the same as 1 but more specific. People use "cloud" to refer to anything connected to a server except you access the server through a service and all the specifics like hostnames and such are abstracted away. Like iCloud for Mac.
I just don't get all the fuss about the Cloud and I'm glad there's someone else out there that scoffs when they hear the term. How is Amazon EC2 "in the cloud"? As far as I can tell its a VPS with some neat scaling features. How is Heroku and other app hosting services "in the cloud"? Again, what I see is VPS hosting with even more abstraction on top of Amazon's abstraction.
Whenever I did freelance development work for people they'd ask if the hosting I offered was in the cloud. First off, why does that matter (well the people who ask for it claim it's better but can't explain why)? Second, what does that even mean? If it means will your site be hosted on a server connected to the Internet then yeah I suppose you can cal it the cloud.
About a year ago I noticed some copy on MediaTemple's page for either their (ve) or (dve) hosting offerings that say something to the effect of "some people call it the cloud but we prefer to think of it as... (some very same logical explanation here)". I was happy to see it because in the end, as far as I can tell there's no difference between the cloud and one or more servers accessible via the Internet.
Maybe to non-technical people having their information and data available across all devices by syncing with a remote server is novel and can only be understood by using the cloud metaphor. To them, the "cloud" is this vague thing on the internets that's all knowing and always available that's the next revolution in computing. To me it's just a client and server connecting and exchanging data. It's been there since I was a kid. The only difference is that now my phone, my car, and even my fridge can connect to it (for some strange reason) but its still just a damn server like we've had since before 56k modems.
Somehow Dropbox is a cloud service but gmail (or hotmail or whatever email provider) is on the Internet according to many. How is that?
--This is exactly it. I think internet is just converging to cloud, because evrything is becoming more hybridiZed. also, datasets are getting larger, so local storage needs to be larger. but thats hard to carry around in a secure way (raid, etc). So, distributed/redundant storage makes sense. But then I want to have mobile access. device is low power/32gb etc. so again, need to rely on distributed storage to enable mobility, not just scale/security. I also think we see this with web-pages themselves. Not pulling from a "server" only, but pulling from all over multiple datasources (ad networks, etc). Easier to just call this distributed storage/hybrid data sourcing "cloud" to avoid wasting thought onn the details (every case different/unknowable). It's a bit of a moving target, then, as to what (if anything) a "cloud" XYZ is.
Reed's Law is arguing that the number of cliques I can participate in on a social network is more important than the number of "friends". The number of cliques tends to grow exponentially in the number of users except:
(i) Social networks are extremely sparse.. the most well connected nodes have 5000 out of hundreds of millions of users. This greatly reduces then number of available cliques (although does still leave it exponential).
(ii) Many of the connections are quite weak.. my interest in a random clique of a social graph in which I'm a member is almost always zero.
(iii) Most cliques that provide value can be extended into other cliques by inviting members, so I may only be interested in maximal cliques, a further significant winnowing.
(iv) Cliques aren't even the greatest representation of this because most groups I participate in on a social network don't have all-to-all friending.
On the whole though I expect the effect of disinterest in most cliques, and improving the interested one's to optimality by extension reduces the cliques I'm actually interested to something far slower than exponential.
Can we rename "the trough of disillusionment" to "the pit of despair"? That's how I've felt when I had products trapped there and it's also a great movie reference.
A reference to Moore's law in the lead paragraphs of
an article in popular media is a useful diagnostic.
If the law is mentioned at all, usually in ~200 words,
I expect the writer to handwave with low signal/noise
about some supposedly latest-and-greatest tech and I
can decide whether to read the rest of the article.
If the writer butchers an explanation of Moore's law,
I can simply skip the rest of the article, confident
that there will be enough other errors to drown out
any surviving signal.
> (number of transistors per chip doubles every X amount of time)
That's closer, but still not there. Moore was more specific: He talked about the chip that currently has the lowest price per transistor.
What really amazes me, is that Moore's law works backwards, too, for much longer periods than Moore had fitted the data to. You need to relax the definition of transistor somewhat, though. But punch-card looms still fit the pattern.
On a tangent: I wonder how long exponential progress in falling prices will last in the fields of genome sequencing and synthesis. (I've also read that rechargeable batteries currently improve by around 8% a year. I don't remember by which metric exactly, though.)
The last is not. Reed's law is a extremely optimistic version of Metcalfe's law which gives clearly nonsensical results. And Metcalfe's law in turn seems to be overly optimistic. See http://spectrum.ieee.org/computing/networks/metcalfes-law-is... for evidence that the real scaling law tends to be more like n log(n). (See http://www.dtc.umn.edu/~odlyzko/doc/metcalfe.pdf for several other lines of argument leading to the same result.)
Note that when I say "tends to be" I mean that, depending on the details of a social network, the scaling law can differ. In particular a network that relies on relative strangers having interactions with other relative strangers, as happens with eBay and Airbnb, is going to scale much closer to Metcalfe's Law. By contrast one that depends on developing a circle of friends, such as happens with Facebook, the scaling law will be closer to n log(n).
Disclaimer, I have a bias here since I'm one of the co-authors of the n log(n) law. (Andrew Odlyzko did all of the work. And then we found out that Bob Briscoe had independently arrived at the same conclusion based on data that he had access to at British Telecom.)