Computing is a field where integration matters a lot and bottlenecks (many of them in the human-machine interface) are often not even recognised as such, but seen as "the way things are". I mean there are a lot of good ideas that nobody has found the way to implement with enough usability so people wants to use them. Was YouTube a new idea when it "exploded"? I don't know, but I believe that it represents an innovation, maybe not in the idea, but in the right way to implement it. I guess there are a lot of innovative ideas in the "implementation pending" state.
"Implementation pending" is a great way to put it. There is a lull between the time an idea can be conceived, it's feasibility proven, and it spreading into the modern zeitgeist. I feel that at this particular point, we know that there are lots of ideas that can be implemented, but aren't, especially in the mobile space (hence, the oligopoly reference in my post). Other continents have disseminated many mobile innovations into the public before the U.S. has in the last ten years.
Just on the YouTube example, it was more an that its time had come, and YouTube just happened to be the one who got popular.
The factors included: multimedia-capable PCs, broadband and cell-phone cameras. All driven by Moore's Law.
A great deal of compsci seems to be like this (even including new ways of multiplying numbers) - not surprising, since Moore's Law dwarfs all other forces. No other industry has anything anywhere near comparable.
This can be used predictively: what obviously great but ridiculously infeasible idea will Moore's Law soon make feasible?
Examples: spherical TV (you sit inside); golf-swing analysis by brute-force matching with 3D recordings of every possible swing (to arbitrary resolution); literal recording and playback of 3D cerebral cortex activity (to arbitrary precision).
Displays are tricky. A 1280x960 is four times the 640x480 that was typical some years ago. We need more memory, more buses bandwidth, bigger and faster everything in different levels (without talking about colour depth), so I'd say that the power needed would be 20x or 30x. My gut feeling is that displays only have consumed Moore's gift for the last ten years.
> We must bridge the gap between the techies and the rest of the population.
This is sooooo important. Blogs existed before blogging platforms; it just was too hard for non-techies to do. Look how long email was around before AOL made it easy and fun. Wikipedia was possible before the wiki had been invented, but the wiki made it more accessible for everyone.
Sure, people are gaining competence in computing, but we need to make things easier to use or consume as well! There's already a lot of great CS/technology out there that has so much potential if we can get more people to embrace it. It's half on people to adopt the new technology, but half on use to make the technology easily adoptable.
That second link is pretty good as a rant, but I don't find it very good as an argument.
The whole argument about macros as a replacement for "static higher order programming," which is supposedly better than "dynamic higher order programming," which evidently nobody needs, was lost on me. He also seems to be saying, though, that "dynamic higher order programming" does not include closures, so I really don't know what it is. If a closure is anything, it is certainly a function created at runtime.
He also seems to say a function composition operator is obviously bad, without any justification. I guess if justification were required, it wouldn't be obvious. I don't see the clear advantage of carrying around a data structure with multiple functions and a third function that composes them, versus just using a compose operator. I suppose it's a little more opaque, but that is often a good thing (a kind of encapsulation).
"Of course, it may be slightly easier to optimize the result of dynamic code generation"
But maybe it's a lot easier? Can we get a huge speedup, or just a little one? I have no idea, because he doesn't say, and it's not even clear if he knows.
"Whatever problem we are solving by constructing h(x), it is probably better solved by a macro, which would perform the same composition statically."
I highly doubt any serious Lisp programmer agrees with this statement. Dynamically constructing functions at runtime is an extremely common lisp (pun unintended, but fortuitous) technique.
I have not programmed in Haskell, but reading this gives me no legitimate reasons to think it is a bad language, let alone an indictment on CS research.
Without having read his other posts it wouldn't be quite as obvious, but the author of the second link (at least in his blog persona of 'Mencius Moldbug') is legitimately insane.
His writings are crafted to shock the hell out of any socialized person, no matter their world-view. He actively despises representative democracy, functional bureaucracies, and anyone advocating policy (the state department, the NYT, NGOs, etc.). You could denounce him as a fascist libertarian, but neither of those groups would have any idea what to do with him either.
Unfortunately you'll have to read several hundred thousand words of blog-essays to even begin to understand him (and you thought Yegge was a bloviator!).
He's not insane. He's an anarchist.(Or seem to be one)
You would do well to understand an anarcho-capitalist's position regarding the incentive of government insitutions versus market based insitutions. Than it become less shocking to read his work because well, you understood where he's coming from.
As for me, I am an anarcho-capitalist myself so I am not shocked. Figures.
After reading most of his archives, I'm still trying to decide whether he's insane, or if he's sane and it's everyone else who is insane. More and more I'm thinking the latter. Perhaps I need to disconnect myself from the internet for a while :-)
