Is this going to be so desirable in a datacentre of all places where power consumption and cooling are already a significant portion of a system's ownership costs?
The answer is yes.
Because the critical considerations are compute efficiency (wallop per watt) and compute density.
Escalating wallop density is ultimately going to cost you on the TDP front, except as compared to free lunch.
A = 2 x (1 platform + 1 CPU) B = 1 x (1 platform + 1 turbo CPU)
Suppose turbo CPU is 50% hotter, 100% more powerful.
Or perhaps you've swallowed the SOC blue pill in a single gulp, and think that the mainboard is just a bunch of fiberglass, epoxy, plastic restraining clips, and copper artwork.
The mighty 6" between your two CPU sockets doesn't come for free, either. (Also known as mighty slow, as in ye mighty 6" of the 1 ns electrical round tip.)
The classic Lewis-Stalnaker semantics for counterfactuals captures that Sobel sequences are consistent sequences, for example:
If Sophie had gone to the parade, she would have seen Pedro dance.
But if Sophie had gone to the parade and been stuck behind someone tall, she would not have seen Pedro dance.
But reverse a sequence like this one and it no longer sounds so good, which is surprising on the classic semantics.
This observation motivated Kai von Fintel (2001) and Thony Gillies (2007) to propose dynamic semantic accounts of counterfactual conditionals.
Subsequently, Sarah Moss (2012) defended the classic semantics against the charge that it need be abandoned in the face of these order effects, arguing that the infelicity of the reverse sequences is pragmatic.
I argue that both accounts are ultimately untenable, but each account has strengths. Seeing what works and what doesn't in each account points the way to the right positive view.
With this in mind, I defend a contextualist account of counterfactuals that takes conversational relevance to play a central role.
What a person "would have preferred" tends to scamper down this rabbit hole in a big hurry.
Jump, jump, you should follow.
Better down a dark rabbit hole than standing in the bright sun boxing with your own shadow.
To give people the power to build community and bring the world closer together.
First problem:
Absence makes the heart grow fonder.
Second problem:
Tinfoil is the great equalizer.
It allows those who really know very little about how the world works (hint: it's complicated) to put up a big front. Any steep ignorance gradient confined to a narrow social context is going to generate a tinfoil IED (improvised exposition device).
We used to blame Alzheimer's disease on leached aluminum. Boy, were we thinking small.
Simultaneous interpreters normally work in teams of two per booth, taking turns in shifts of about 30 minutes each for a maximum of about three hours at a time, which has been found to be the maximum average time during which the necessary concentration and accuracy can be sustained.
For some reason, this came up in a machine learning resource I consumed recently.
Of course, this job is likely to be impacted fairly significantly, if the computers ever get to the really high level of accuracy required to pass the SALT hurdle.
(Any experienced U.N. translator would get that reference instantaneously.)
It will always be more efficient to have the higher quality worker work 40 hours.
That's why the Edmonton Oilers played Connor McDavid for almost 60 minutes per-game over the last season, or more if the game went into OT.
Really, you've got six guys dressed to actually play, four more guys to penalty kill (a specialized skill), and another dozen guys dressed to briefly jump over the boards and send a physical message, as the situation warrants—which seems to be the case, almost invariably, three shifts out of four.
Weird.
That's a lot more "messaging" than the self-evident efficiency curve would seem to suggest.
Depends on whether you render AI as "artificial intelligence" (dumb and tired) or "automagic induction" (smart and wired).
Automatic induction is rocking out, lately, with important applications constructed using general purpose learning algorithms, mounds of data, and very little hand-crafted (expensive) feature logic.
But if you're satisfied spending the rest of your life griping about scant progress at clearing the One Ridiculously High Bar to Rule Them All (1950s-style), go right ahead.
Sub-second trading granularity just isn't adding anything meaningful, but it causes crazy distorted and unfair trading practices. Something like shuffling transactions over a larger timescale that is more fair.
It's extremely hard to create a batch clearance protocol (say, once per minute) where you don't create an information hazard where firm A arranges to receive a vital public stock press release a 0.1 ms before the cutoff time, while it's competitor receives the same press release 0.1 ms after the cutoff time. You just have to pick the right city to post the first formal announcement.
Trust me, with this kind of trading advantage available, Bitcoin would flow into the dark wallets of press release enter-key jockeys everywhere.
Okay, you have a better idea. So you create a worldwide publication system that makes the same information available to everyone 10 s before the trade block, after which even Twitter falls silent. Now everyone bidding is on a level information playing field. (Hey there, Mr Greedy Stockbroker, no talking out of school using spread spectrum!) Maybe you settle a global trade auction once per hour, prior to which all new information on the planet that makes the cut is bundled up into some kind of giant ZFS Merkel tree, and everyone gets a copy. Then, for 10 s before the hour, all global fiber and radio goes dark until everyone places their fair bids, to be resolved by the One True Settlement Pass.
Sound like a plan?
HFT truly is a pox on humanity, but unfortunately, the pox has a billion year precedent: organisms have always competed for having the fastest nervous systems, and figuring out a way to gain proximity on the vital and timely survival information.
At the end of the day, HFT is just biology pursued by other means.
But with some peculiar twists.
Imagine how—in this New World Order—the Pompeii office places some extremely shrewd bets in the microseconds before it ceases to exist.
I've so far only managed to come up with one proper solution: build a Ringworld. A Ringworld where all news is effectively local news (so long as walls perform). A Fukushima incident takes out the entire conflict metal supply in Scorpio quadrant? Surely the localized market blip is hard to detect a mere 5 light-seconds upring or downring.
Some of the most interesting features of the Ringworld economy are dictated less by its size, than it's fundamentally linear macro-scale geometry.
Now, there would still be some competition to create the fastest possible communication network, Ringworld-wide ("wide" on Ringworld generally means "trivial in size", but we'll ignore this). The fastest diametric system probably involves shooting polarized neutrinos straight through the sun. If macro-HFT nanoseconds matter enough to you, surely you'll figure this out.
As another example, the Pentagon will be freed from a requirement that it file a report every time a small business vendor is paid, a task that consumed some 1,200 man-hours every year.
The Distributed Aperture System is one of the primary sensors feeding the displays to the infamous $600,000 helmet system, and it is also failing to live up to the hype.
So if we amortise the JSF program over 40 years, the $1.4 trillion outlay / pork gravy works out to $1100/s (more than even Eliza "meth" Millipede can make stuffing fliers at home).
Is 1200 hours/year of accounting oversight on a relatively small financial leak unjustified?
Personally, I'd crank some numbers before jumping to a hasty conclusion, because this is the ultimate haemorrhage in all of recorded history.
I happen to mostly take complaints about the information system with a grain of salt. Integration problems are hard, and things will probably improve with a combination of time, experience, and more $$$ guzzling everywhere (perhaps in some cases to good effect).
The F-35 lost repeatedly in air-to-air maneuvering despite the fact that the test was rigged in its favor because the F-16 employed was the heavier two-seater version and was further loaded down with heavy, drag-inducing KFC chicken buckets to hinder its maneuverability.
F-35 boosters argue that the plane's low radar signature will keep it out of WVR situations, but the history of air combat is that WVR engagements cannot be avoided altogether.
Incompetent epsilon-signature airframe, abetted with movable mission goalposts, have a far worse long-term prognosis.
But no, apparently the problem here is too much picayune cost control.
Personally, I'd crunch some serious numbers before supporting that assertion in any direction.
The least distorting tax is a small tax, and the smallest tax is a broad tax. I'm in favour of taxing everything by a mostly equal, small amount. Less discretion, less political dickering.
People do realize that the taxation system is a waterbed...
The internet infrastructure works whether or not you send bits over it, it doesn't wear out any faster nor does it need any more maintenance because a bit was sent.
Apparently you have yet to meet our current and future generations of non-volatile memory.
A network computer whose merest operational logfile I am not worthy to exfiltrate—and yet I will design it for you.
I constantly marvel at how Douglas Adams got everything deeply right, whereas Clifford Stoll, not so much.
Today an eight-foot-long Keuffel & Esser slide rule hangs on my wall. Once used to teach the mysteries of analog calculation to budding physics students, it harkens back to a day when every scientist was expected to be slide-rule literate. Now a surfboard-size wall hanging, it serves as an icon of computational obsolescence. Late at night, when the house is still, it exchanges whispers with my Pentium. "Watch out," it cautions the microprocessor. "You never know when you're paving the way for your own successor."
Wise, slightly overclocked Pentium: If I'm not paving the way for my successor, it can only be due to my FDIV bug.
Keuffel & Esser: This one time, at band camp, I fell into a bath tub.
Pentium: A likely story. I might be mathematically challenged, but look at you, you're eight-feet long!
Keuffel & Esser: True that. It was just my epsilon end that became unreliable.
No, this situation arose because the government regulates the airwaves, thereby limiting the number of companies which can compete in the cellular market. With few competitors, it's very easy for them to coincidentally decide to lock their phones.
Power abhors competition. It doesn't matter what form this power takes. When you have strong government, government gets blamed. Where you lack strong government, factional violence gets blamed.
There's a principle in computer science that you can solve any problem by adding another layer of indirection, except for too many layers of indirection.
There's a principle in politics that you have nothing to fear but fear itself.
There ought to be an economic principle that you can solve any problem by enabling more competition, except for the competition to destroy competition.
Power abhors competition.
Unfortunately, economists waste too much of their brain power on ideology and fail to spot the obvious.
Now nothing makes government power intrinsically benign, but it is easier to point to. (What does Wall Street contribute, exactly, that benefits the economy enough to justify raking in 1/6 off all corporate profit in America?)
In a country with strong public institutions, being easier to point to can make government more benign than diffuse, libertarian oligarchy.
Hence libertarianism abhors strong public institutions.
The bottom line here is that competition itself is subject to systems theoretic scarcity constraints. These are especially hard to point to, but they nevertheless exist.
The next Hayek will be the guy who figures out how to write this down in compelling equations.
And no, that's a rhetorical question. There is no additional content in the article which would add anything remotely of value.
I have trouble making sense of the news cycle making heavy use of all three monitors on my desk.
I'm almost always Googling any significant name or place or context I'm not familiar with, and cross-referencing one article against another, or checking out whether a source is sane by seeking out other material associated with same.
I guarantee you that my understanding of the news differs from yours, and my broad modality would not survive if I consumed news on a crappy little phone.
I consume almost nothing on my phone unless I'm forced to by circumstance, because my standard of context is too large to fit the device. I'm also a guy who remembers 80x25 as an upgrade from a world of extreme pain. Once again I find myself at 40x25, but this time it's inches and I'm never going back.
Base rate for humans running around with their hair on fire over some anomaly that later proved to be something entirely different: ludicrous speed.
Base rate for humans inventing various super-beings and spirits and pseudo-species out of whole cloth: surely nothing to sneeze at.
Number of star systems beaming fake intergalactic news at ever higher levels of transmission power until crystal diodes are exploding everywhere: zero (give or take).
Furtive space aliens on a "two hour cruise" who chanced into near-earth proximity at any point in the last 1000 years: well, you just never know.
This kind of weird name is given to pie-eyed future technology projects so that when the dust settles no-one really knows precisely what didn't pan out.
Because odds are, they're going to have to fund this again—with an inkling of clue & a vaguely comprehensible name—before this twinkle finally deposits a nugget, third time lucky.
AI gets to the point were it solves a set of previously unsolvable problems, the algorithms are then researched and better non-AI solutions are then used to solve the same problems. Then AI falls out of fashion for a while and computer power increases thanks to Moore's law. Then it all repeats.
This old story is such a crock.
I highly recommend the following to anyone who wants a different perspective on modern ML:
This is plain old numerical methods, optimization, and search viewed through a Bayesian inference filter. I would never have termed any of this "artificial intelligence".
It took the recent large advances in unsupervised learning, the kitty classifier (and progeny), and the LSTM machine translation models to finally justify rethinking academic labels. Programs like SHRDLU from 1968 were perhaps explorations in AI, if our baby-step microscope is especially well focused. But this was closer to natural philosophy than what later became physics. Even our shiny new LSTM language models remain weirdly proximal to Searle's Chinese room. What have we really learned from watching our machines learn? Not a whole damn lot.
I'd nominate a term such as I-cubed: inexplicable inductive inference, or perhaps MIII: massively inexplicable inductive inference.
Even so impeded with an appropriate name, MIII is pretty mind-blowing. But it still ain't AI. It might be a viable building block to proceed in that direction, sooner rather than later, as we begin to erect dynamical systems upon this foundation. To drive the point home, it remains way overblown to call it MIIR: massive inexplicable inductive reasoning.
"Before AlphaGo, much of the fundamental games and machine learning research was done here," Muller wrote in an email. "If you look through the references list of the AlphaGo paper in the journal Nature, over 40% of these references have a University of Alberta (co-)author. Then, DeepMind greatly surpassed all of these previous efforts with their new ideas."
I haven't waded through this yet, but I suspect even the vaunted AlphaGo has a backbone of techniques that I personally wouldn't have classed as "AI" (or even AI-ish) by my own standards.
For decades, the big idea in AI was supposed to be recursion. Perhaps human language is recursive in theory, but it's only barely recursive in practice (nest more than three levels, your accurately attentive audience grows thin). Winograd is not completely wrong about this, but my long suspicion is that recursion is not going to enter the AI building through the ground floor.
Lately, we really have hit home runs with distributed representation, and to a lesser degree with convolutional image recognition. These are actual AI-ish ideas. However, two solid take-home techniques do not a field make.
Here's another possible intermediate term: generalized gradient exploitation (GGE). Plus there's tons of great mathematics about overfitting and regularization. But should we really call all this math "AI"?
In practice what AI ends up meaning is "look ma, no code!" Hey, we just built an impressive system without hiring rooms full of code monkeys, so we must be doing something right.
AI is not the moving target of lore. It's mainly our long AI pretension that fits the bill.
There is little question, Intel is likely none too pleased with it and PC OEM heavyweights Lenovo, Hewlett-Packard and ASUS have also signed-on to deliver Windows 10 notebooks and 2-in-1 convertibles powered by Qualcomm.
Take that, every English teacher you've ever had from grade 3 onwards.
In another strip, after stating that he could not identify Plymouth Rock, lest it "compromise our agents in the field," Calvin cheerfully remarks, "I understand my tests are popular reading in the teacher's lounge".
There's the difference. Calvin is fighting the good fight. Whereas the above sentence is clown-car laser water-canon own-goal toward the cause of youthful creative autonomy.
Why stop at 2 (if duopolies are also too bad)? Why not 3 or a 100 or a gazillion?
Just a tip if you're stymied: if you start with your shoe, your ass is on the other side of your knee.
Just another tip if you're still stymied: your kneecap is a hard bump roughly halfway up your leg.
Okay, one last tip if you're almost beyond salvation: I wouldn't spend a lot of time feeling around inside your socks or your pockets to locate your kneecap.
Why not 3 or a 100 or a gazillion?
Because you won't locate your kneecap fumbling around inside your socks or your pockets.
Middle ground, present in a leg to stand on, presumably somewhere near you.
There is one rational explanation that, if true, would totally justify this rapid increase in price across some of the major tulip germ lines. And that is, maybe these tulips are actually worth these high prices, and maybe even worth many times more than that at which they are currently trading.
Am I the dullest knife in the drawer?
Because, seriously, I can't tell the difference between these two arguments.
I dunno about of grid guy. That kinda person is rare and interesting enough they might get some physical surveillance.
I think people are in part attracted to this subject for all the opportunities it provides to masturbate over the presumptively infinite dark budget—flip side of one mad guy conducting Symphony of a Thousand on secret volcanic island.
Read my lips: mass surveillance.
As in Henry Ford. As in McDonald's. That's the whole point of our present-day military-industrial economic order.
I used to ask myself, why are all the crazy people trying to make bioreactors to generate ethanol, or some other carbon-based fuel, when they could be bio-reacting a nitrogen-based fertilizer instead (the quantities involved are more realistic). But perhaps the NSA doesn't want random people halfway off the grid to possess this particular biotechnology.
So perhaps in this case they might actually eff themselves to roll a damn-expensive invisible, black truck to surveil some marginal crank with a gene printer apparently subsisting on juniper berries in a long lost valley.
You know what?
I'm barely biting on my own rabid edge case.
The NSA is simply too damn cheap. The most they're gonna roll is a custom regex.
It's not rocket science people. The guy won the election. That means there are over a hundred million people in the USA who want him there and want to hear what he has to say.
I've probably seen more actual footage of Trump on the stump than any other politician alive.
He said it loud, and he said it often.
Furthermore, people vote to give the candidate the opportunity to run the country (often holding their noses while doing so). It's a prospective act.
So, yes, back on election day, 100 million Americans wanted to audition Trump as president more than Hillary, in an election widely depicted as a nose-gripping low bar.
"We did not evolve from a single cradle of mankind somewhere in East Africa," said Phillipp Gunz, a paleoanthropologist at the Max Planck Institute for Evolutionary Anthropology in Liepzig, Germany.
So one individual decides to make a long, solo trek to toss an alien, divisive artefact off the edge of the world, and now our entire "cradle" theory is shot, all because some magnetically addled frigatebird dropped a clam shell of rancor right down the maw of some inland rift valley.
Most any scientist will tell you that the entire process of publication and review is political.
And dinner is mostly made of water. But is that the key thing?
The point here is not that science escaped politics, but that it somehow progresses (over the long run) nevertheless.
This has long been one of my complaints about climate science. The window to fully separate science from politics has historically been 50 to 100 years. As such, the "science" of imminent ruin is not a valid, established tradition.
And yet, here we are, forced to make decisions at climate gun point, on a still 50% political consensus that this gun even exists, which is surely but slowly turning into what I count as real science. When the dust settles, the present consensus will not necessarily endure as the story we presently tell ourselves in its primary outline.
* The Century Of The Self * Ken Burns; for us, most recently, it was The West which we found surprisingly informative, once we got used to the dead-slow pacing * James Burke
Both on my TODO list already: * David Attenborough * Louis Theroux; so far only Louis and the Nazis which was interesting, but also imperfect
Not yet seen here: * Glenn Gould: The Russian Journey
It was on YouTube in early 2016, but has since flat-mouthed (flat != level). Here's one remarkable passage:
I will tell you something that no-one knows, because it is from my intimate conversion with Sviatoslav Richter. All of sudden he says "you know, I can play Bach as well as Gould. But you know why I don't play as well as he does? Because I would have to work so hard to play that way. This is the genius of Glenn Gould."
David Hoffman's documentary on John von Neumann was also on YouTube (as of early 2015) and has also subsequently flat-mouthed. Old, old, old, but I found it amazing.
Another thing, some of the molecular animations that started to come out in the 2007 time frame are mind blowing. (Three names from my notes are Janet Iwasa, Drew Berry, and David Bolinsky, but I don't have them attached to specific clips.)
Just to get ahead of this, for podcasts, one of my favourites episodes recently was Talking Machines interviewing Douglas Eck on generative art.
The "Alpha Gal" episode on RadioLab was hugely entertaining (and useful for terrifying your meat-loving relatives).
Finally, I'm also just loving the new (not many episodes yet) More Perfect about the U.S. Supreme Court which recently spun out of RadioLab.
If you find all that too entertaining, Bus 174 and Darwin's Nightmare (both quasi documentaries) will cure you in short order (mesmerized != amused).
And, oh yeah, one can do far worse than the various documentaries of Werner Herzog.
Slashdot Mirror
The answer is yes.
Because the critical considerations are compute efficiency (wallop per watt) and compute density.
Escalating wallop density is ultimately going to cost you on the TDP front, except as compared to free lunch.
A = 2 x (1 platform + 1 CPU)
B = 1 x (1 platform + 1 turbo CPU)
Suppose turbo CPU is 50% hotter, 100% more powerful.
Or perhaps you've swallowed the SOC blue pill in a single gulp, and think that the mainboard is just a bunch of fiberglass, epoxy, plastic restraining clips, and copper artwork.
The mighty 6" between your two CPU sockets doesn't come for free, either. (Also known as mighty slow, as in ye mighty 6" of the 1 ns electrical round tip.)
You seem to suffer from a complete and utter inability to navigate counterfactual statements.
Counterfactual Discourse in Context — January 2017
What a person "would have preferred" tends to scamper down this rabbit hole in a big hurry.
Jump, jump, you should follow.
Better down a dark rabbit hole than standing in the bright sun boxing with your own shadow.
First problem:
Absence makes the heart grow fonder.
Second problem:
Tinfoil is the great equalizer.
It allows those who really know very little about how the world works (hint: it's complicated) to put up a big front. Any steep ignorance gradient confined to a narrow social context is going to generate a tinfoil IED (improvised exposition device).
We used to blame Alzheimer's disease on leached aluminum. Boy, were we thinking small.
Just in case you think I was only being facetious.
Simultaneous Interpreting: Some Frequently Asked Questions!
For some reason, this came up in a machine learning resource I consumed recently.
Of course, this job is likely to be impacted fairly significantly, if the computers ever get to the really high level of accuracy required to pass the SALT hurdle.
(Any experienced U.N. translator would get that reference instantaneously.)
That's why the Edmonton Oilers played Connor McDavid for almost 60 minutes per-game over the last season, or more if the game went into OT.
Really, you've got six guys dressed to actually play, four more guys to penalty kill (a specialized skill), and another dozen guys dressed to briefly jump over the boards and send a physical message, as the situation warrants—which seems to be the case, almost invariably, three shifts out of four.
Weird.
That's a lot more "messaging" than the self-evident efficiency curve would seem to suggest.
Briefly, your raw curve was almost so convincing.
Depends on whether you render AI as "artificial intelligence" (dumb and tired) or "automagic induction" (smart and wired).
Automatic induction is rocking out, lately, with important applications constructed using general purpose learning algorithms, mounds of data, and very little hand-crafted (expensive) feature logic.
Feature engineering is pretty much a dead career already.
But if you're satisfied spending the rest of your life griping about scant progress at clearing the One Ridiculously High Bar to Rule Them All (1950s-style), go right ahead.
It's extremely hard to create a batch clearance protocol (say, once per minute) where you don't create an information hazard where firm A arranges to receive a vital public stock press release a 0.1 ms before the cutoff time, while it's competitor receives the same press release 0.1 ms after the cutoff time. You just have to pick the right city to post the first formal announcement.
Trust me, with this kind of trading advantage available, Bitcoin would flow into the dark wallets of press release enter-key jockeys everywhere.
Okay, you have a better idea. So you create a worldwide publication system that makes the same information available to everyone 10 s before the trade block, after which even Twitter falls silent. Now everyone bidding is on a level information playing field. (Hey there, Mr Greedy Stockbroker, no talking out of school using spread spectrum!) Maybe you settle a global trade auction once per hour, prior to which all new information on the planet that makes the cut is bundled up into some kind of giant ZFS Merkel tree, and everyone gets a copy. Then, for 10 s before the hour, all global fiber and radio goes dark until everyone places their fair bids, to be resolved by the One True Settlement Pass.
Sound like a plan?
HFT truly is a pox on humanity, but unfortunately, the pox has a billion year precedent: organisms have always competed for having the fastest nervous systems, and figuring out a way to gain proximity on the vital and timely survival information.
At the end of the day, HFT is just biology pursued by other means.
But with some peculiar twists.
Imagine how—in this New World Order—the Pompeii office places some extremely shrewd bets in the microseconds before it ceases to exist.
I've so far only managed to come up with one proper solution: build a Ringworld. A Ringworld where all news is effectively local news (so long as walls perform). A Fukushima incident takes out the entire conflict metal supply in Scorpio quadrant? Surely the localized market blip is hard to detect a mere 5 light-seconds upring or downring.
Some of the most interesting features of the Ringworld economy are dictated less by its size, than it's fundamentally linear macro-scale geometry.
Now, there would still be some competition to create the fastest possible communication network, Ringworld-wide ("wide" on Ringworld generally means "trivial in size", but we'll ignore this). The fastest diametric system probably involves shooting polarized neutrinos straight through the sun. If macro-HFT nanoseconds matter enough to you, surely you'll figure this out.
Please reformulate that statement as a viable Bayesian prior.
Inquiring minds want to know.
For extra points, precisely where does this "seems" originate, and, most crucially, does it resemble a starfish?
Sounds like a lot, I guess.
The F-35 Is a $1.4 Trillion Dollar National Disaster — April 2017
So if we amortise the JSF program over 40 years, the $1.4 trillion outlay / pork gravy works out to $1100/s (more than even Eliza "meth" Millipede can make stuffing fliers at home).
Is 1200 hours/year of accounting oversight on a relatively small financial leak unjustified?
Personally, I'd crank some numbers before jumping to a hasty conclusion, because this is the ultimate haemorrhage in all of recorded history.
I happen to mostly take complaints about the information system with a grain of salt. Integration problems are hard, and things will probably improve with a combination of time, experience, and more $$$ guzzling everywhere (perhaps in some cases to good effect).
Incompetent epsilon-signature airframe, abetted with movable mission goalposts, have a far worse long-term prognosis.
But no, apparently the problem here is too much picayune cost control.
Personally, I'd crunch some serious numbers before supporting that assertion in any direction.
The least distorting tax is a small tax, and the smallest tax is a broad tax. I'm in favour of taxing everything by a mostly equal, small amount. Less discretion, less political dickering.
People do realize that the taxation system is a waterbed ...
Don't they?
Apparently you have yet to meet our current and future generations of non-volatile memory.
I constantly marvel at how Douglas Adams got everything deeply right, whereas Clifford Stoll, not so much.
When Slide Rules Ruled — 2006
Wise, slightly overclocked Pentium: If I'm not paving the way for my successor, it can only be due to my FDIV bug.
Keuffel & Esser: This one time, at band camp, I fell into a bath tub.
Pentium: A likely story. I might be mathematically challenged, but look at you, you're eight-feet long!
Keuffel & Esser: True that. It was just my epsilon end that became unreliable.
Power abhors competition. It doesn't matter what form this power takes. When you have strong government, government gets blamed. Where you lack strong government, factional violence gets blamed.
There's a principle in computer science that you can solve any problem by adding another layer of indirection, except for too many layers of indirection.
There's a principle in politics that you have nothing to fear but fear itself.
There ought to be an economic principle that you can solve any problem by enabling more competition, except for the competition to destroy competition.
Power abhors competition.
Unfortunately, economists waste too much of their brain power on ideology and fail to spot the obvious.
Now nothing makes government power intrinsically benign, but it is easier to point to. (What does Wall Street contribute, exactly, that benefits the economy enough to justify raking in 1/6 off all corporate profit in America?)
In a country with strong public institutions, being easier to point to can make government more benign than diffuse, libertarian oligarchy.
Hence libertarianism abhors strong public institutions.
The bottom line here is that competition itself is subject to systems theoretic scarcity constraints. These are especially hard to point to, but they nevertheless exist.
The next Hayek will be the guy who figures out how to write this down in compelling equations.
I have trouble making sense of the news cycle making heavy use of all three monitors on my desk.
I'm almost always Googling any significant name or place or context I'm not familiar with, and cross-referencing one article against another, or checking out whether a source is sane by seeking out other material associated with same.
The medium is the message
I guarantee you that my understanding of the news differs from yours, and my broad modality would not survive if I consumed news on a crappy little phone.
I consume almost nothing on my phone unless I'm forced to by circumstance, because my standard of context is too large to fit the device. I'm also a guy who remembers 80x25 as an upgrade from a world of extreme pain. Once again I find myself at 40x25, but this time it's inches and I'm never going back.
Base rate for humans running around with their hair on fire over some anomaly that later proved to be something entirely different: ludicrous speed.
Base rate for humans inventing various super-beings and spirits and pseudo-species out of whole cloth: surely nothing to sneeze at.
Number of star systems beaming fake intergalactic news at ever higher levels of transmission power until crystal diodes are exploding everywhere: zero (give or take).
Furtive space aliens on a "two hour cruise" who chanced into near-earth proximity at any point in the last 1000 years: well, you just never know.
News at 11.
This kind of weird name is given to pie-eyed future technology projects so that when the dust settles no-one really knows precisely what didn't pan out.
Because odds are, they're going to have to fund this again—with an inkling of clue & a vaguely comprehensible name—before this twinkle finally deposits a nugget, third time lucky.
This old story is such a crock.
I highly recommend the following to anyone who wants a different perspective on modern ML:
* Talking Machines: Remembering David MacKay with Philipp Hennig — 21 April 2016
* Probabilistic-Numerics.org
This is plain old numerical methods, optimization, and search viewed through a Bayesian inference filter. I would never have termed any of this "artificial intelligence".
It took the recent large advances in unsupervised learning, the kitty classifier (and progeny), and the LSTM machine translation models to finally justify rethinking academic labels. Programs like SHRDLU from 1968 were perhaps explorations in AI, if our baby-step microscope is especially well focused. But this was closer to natural philosophy than what later became physics. Even our shiny new LSTM language models remain weirdly proximal to Searle's Chinese room. What have we really learned from watching our machines learn? Not a whole damn lot.
I'd nominate a term such as I-cubed: inexplicable inductive inference, or perhaps MIII: massively inexplicable inductive inference.
Even so impeded with an appropriate name, MIII is pretty mind-blowing. But it still ain't AI. It might be a viable building block to proceed in that direction, sooner rather than later, as we begin to erect dynamical systems upon this foundation. To drive the point home, it remains way overblown to call it MIIR: massive inexplicable inductive reasoning.
An Alberta AlphaGo Pioneer Is in China to Watch the AI Wallop Human Opponents
I haven't waded through this yet, but I suspect even the vaunted AlphaGo has a backbone of techniques that I personally wouldn't have classed as "AI" (or even AI-ish) by my own standards.
For decades, the big idea in AI was supposed to be recursion. Perhaps human language is recursive in theory, but it's only barely recursive in practice (nest more than three levels, your accurately attentive audience grows thin). Winograd is not completely wrong about this, but my long suspicion is that recursion is not going to enter the AI building through the ground floor.
Lately, we really have hit home runs with distributed representation, and to a lesser degree with convolutional image recognition. These are actual AI-ish ideas. However, two solid take-home techniques do not a field make.
Here's another possible intermediate term: generalized gradient exploitation (GGE). Plus there's tons of great mathematics about overfitting and regularization. But should we really call all this math "AI"?
In practice what AI ends up meaning is "look ma, no code!" Hey, we just built an impressive system without hiring rooms full of code monkeys, so we must be doing something right.
AI is not the moving target of lore. It's mainly our long AI pretension that fits the bill.
Take that, every English teacher you've ever had from grade 3 onwards.
There's the difference. Calvin is fighting the good fight. Whereas the above sentence is clown-car laser water-canon own-goal toward the cause of youthful creative autonomy.
Advantage, Miss Wormwood.
Just a tip if you're stymied: if you start with your shoe, your ass is on the other side of your knee.
Just another tip if you're still stymied: your kneecap is a hard bump roughly halfway up your leg.
Okay, one last tip if you're almost beyond salvation: I wouldn't spend a lot of time feeling around inside your socks or your pockets to locate your kneecap.
Because you won't locate your kneecap fumbling around inside your socks or your pockets.
Middle ground, present in a leg to stand on, presumably somewhere near you.
Am I the dullest knife in the drawer?
Because, seriously, I can't tell the difference between these two arguments.
I think people are in part attracted to this subject for all the opportunities it provides to masturbate over the presumptively infinite dark budget—flip side of one mad guy conducting Symphony of a Thousand on secret volcanic island.
Read my lips: mass surveillance.
As in Henry Ford. As in McDonald's. That's the whole point of our present-day military-industrial economic order.
I used to ask myself, why are all the crazy people trying to make bioreactors to generate ethanol, or some other carbon-based fuel, when they could be bio-reacting a nitrogen-based fertilizer instead (the quantities involved are more realistic). But perhaps the NSA doesn't want random people halfway off the grid to possess this particular biotechnology.
So perhaps in this case they might actually eff themselves to roll a damn-expensive invisible, black truck to surveil some marginal crank with a gene printer apparently subsisting on juniper berries in a long lost valley.
You know what?
I'm barely biting on my own rabid edge case.
The NSA is simply too damn cheap. The most they're gonna roll is a custom regex.
I've probably seen more actual footage of Trump on the stump than any other politician alive.
He said it loud, and he said it often.
Furthermore, people vote to give the candidate the opportunity to run the country (often holding their noses while doing so). It's a prospective act.
So, yes, back on election day, 100 million Americans wanted to audition Trump as president more than Hillary, in an election widely depicted as a nose-gripping low bar.
What do you think was printed on the ballot?
[x] entertain me
[_] Crooked Hillary
No, I don't think that was the actual ballot.
So one individual decides to make a long, solo trek to toss an alien, divisive artefact off the edge of the world, and now our entire "cradle" theory is shot, all because some magnetically addled frigatebird dropped a clam shell of rancor right down the maw of some inland rift valley.
And dinner is mostly made of water. But is that the key thing?
The point here is not that science escaped politics, but that it somehow progresses (over the long run) nevertheless.
This has long been one of my complaints about climate science. The window to fully separate science from politics has historically been 50 to 100 years. As such, the "science" of imminent ruin is not a valid, established tradition.
And yet, here we are, forced to make decisions at climate gun point, on a still 50% political consensus that this gun even exists, which is surely but slowly turning into what I count as real science. When the dust settles, the present consensus will not necessarily endure as the story we presently tell ourselves in its primary outline.
I really dislike science conducted at gun point.
There's a lot here I can second.
* The Century Of The Self
* Ken Burns; for us, most recently, it was The West which we found surprisingly informative, once we got used to the dead-slow pacing
* James Burke
Both on my TODO list already:
* David Attenborough
* Louis Theroux; so far only Louis and the Nazis which was interesting, but also imperfect
Not yet seen here:
* Glenn Gould: The Russian Journey
It was on YouTube in early 2016, but has since flat-mouthed (flat != level). Here's one remarkable passage:
David Hoffman's documentary on John von Neumann was also on YouTube (as of early 2015) and has also subsequently flat-mouthed. Old, old, old, but I found it amazing.
Another thing, some of the molecular animations that started to come out in the 2007 time frame are mind blowing. (Three names from my notes are Janet Iwasa, Drew Berry, and David Bolinsky, but I don't have them attached to specific clips.)
Just to get ahead of this, for podcasts, one of my favourites episodes recently was Talking Machines interviewing Douglas Eck on generative art.
The "Alpha Gal" episode on RadioLab was hugely entertaining (and useful for terrifying your meat-loving relatives).
Finally, I'm also just loving the new (not many episodes yet) More Perfect about the U.S. Supreme Court which recently spun out of RadioLab.
If you find all that too entertaining, Bus 174 and Darwin's Nightmare (both quasi documentaries) will cure you in short order (mesmerized != amused).
And, oh yeah, one can do far worse than the various documentaries of Werner Herzog.