The title of this post was needlessly provocative, but the bottom line is: If Apple is taking the moral high ground (or something) re data harvesting - is this fundamentally incompatible with the idea of providing a smart AI experience? If they don't collect user data, then they have limited options:
1- provide an algorithmic experience. That's difficult. The "fuzzy logic" machine learning systems everyone else is using exist because writing strict algorithms to surface all the relevant data is hard.
2- provide a crappy experience from their own AI engine trained on limited data.
3- provide a ??? experience by buying someone else's training dataset.
This isn't even an Apple-specific question. In a world where we're talking about data privacy, in MANY fields (calendar/email/browser data is only the tip of the iceberg - consider autonomous driving data for example, including everything every autonomous car captures with every one of its sensors on every drive) - where is the tradeoff between "we want this thing to look smart" vs "we don't want to feed the beast"?
A signal sent from the client that relies on the server side to honor some sort of contract is pointless. Even if there had been industry consensus on "what to do" with DNT, the situation would be more complex than that even in a world full of good actors. For example, GDPR would set limits on how to respond to DNT that differ from the limits set by US legislation. So from the user perspective, the exact effect of DNT would be opaque, it would be basically "turn off as much tracking as is required by the country that hosts the site, or maybe the country the site thinks you're calling from, or maybe the country the site thinks governs your particular account, if you're signed in.
The inherent lack of clarity is bad enough, but it pales in comparison to the real problem, which is that are very few good actors on the other end of the wire, and an end user has no way to scrutinize them. By the time it is divulged that SiteX is illegally ignoring DNT and storing information outside whatever the local law permits, it's too late - the information is already sold to a thousand different data brokers and it is as undeletable as a nude selfie posted publicly to Facebook.
The only meaningful solution is to build the protections into the client side, so that the client is prevented from sending data that can be gathered by the server end. The server cannot, and can never be, trusted by the end-user. It's disappointing that the options we have in browsers (even with extensions) are still relatively coarse. For example, we need the ability to block all active scripting (including that embedded in a page, not just by blocking specific URLs to malware Javascript sources) except for a small whitelist of items critical to the function of the site. We need a way of blocking particular APIs from being accessed by active web content (there is NO reason why a website needs to know my battery level on a mobile device. If there was a reason, it would be a very limited use case that I would only enable for that one particular site. Same principle applies to a lot of the data that's used to fingerprint browsers).
Mmm. I haven't used Azure app stuff, but I'm aware that.. oh at least 20 years ago, the "Visual" languages, especially Visual BASIC, were already claiming some degree of this program-by-manipulating-blocks-and-setting-attributes (I never went beyond playing with it, but a coworker at the time was very enthused with it and in fact went off to start some kind of one-man business around it). And I have worked with some IDEs that let you do a lot of the programming graphically (for example FPGA IDEs that can convert schematic capture to VHDL/Verilog, which you can then edit before synthesis). I'm ambivalent about this latter group, but they're a really specialized example and probably don't generalize well.
For me, it comes down to "is this making code that runs more verifiably, or that can be developed faster", or some other advantage. I haven't yet seen a compelling argument - but I haven't seen proof that it's impossible either, so I continue to wait and see, and in the meantime continue to grovel in C and assembler because that's the fastest way for me to get little personal projects up and running.
Can there be other HMIs that work, and perhaps *in specific sets of constraints and for particular use cases* work better than qwerty? Sure. But that's not what I see when I read the original discussion. Swords at dawn, sir, on the exact semantics of the word "replace".
Perhaps you fall into the masochist class of authors, then. However we appear to be talking in circles. There has yet to be devised an HMI that is as suitable for the entry of bulk text as the qwerty keyboard or a derivative of it. Nothing currently extant is even remotely as capable. Even if a serious alternative were devised and could be demonstrated to be clearly superior, it would have a massive acceptance cliff to climb, given that everyone would still be learning how to use a regular keyboard. Ergo, the only way such a new HMI will have any chance of success is if it coincides with some drastic change in the tasks people perform and the way they communicate, such that the new HMI is clearly superior for entering data into the new communication method. As long as we talk in words composed of strings of letters, the keyboard will remain the best tool for the job.
BTW, as you are now saying "how good could a mobile HMI be for making notes", that is a completely, completely different question from the original question, which was "can the qwerty keyboard be replaced?" (by implication superseded). A replacement for a thing has to exceed its capabilities in all use cases that matter. Yes, as I indicated, I too use a mobile tool where it's all I have available. But it's not a replacement for the real tool; it's just "the tool you have is better than no tool at all" and for me an electronic text entry tool with a loathsome entry method is slightly more useful than a moleskine because although I enjoy writing in a notebook much more than I enjoy poking about on a touchscreen, having my notes born digital so I can email them to myself and a) never lose them, and b) cut and paste directly into my WP.
In much the same way, a Swiss Army knife can be used to do a lot of emergency repairs or even surgery. And if it's all you've got, you use it. But you get better results, more consistently, and with a better user experience, when you're using a real set of tools or surgical equipment.
The BlackBerry sported a physical (though ergonomically hobbled by the constraints of its form factor) keyboard to solve the same problem - "how to enter text". It wasn't a radically new HMI in the way a touchscreen or a voice recognition is; it was a physically miniaturized version of our old qwerty friend. I too have done proofreading and copyediting, and even some small redrafts of articles I was working on, on a BlackBerry - but only because it was the only tool I had available to me. (At the time I spent a lot of time sitting in bars that didn't approve of laptops. I could nurse a drink and work on my BB as long as I liked though).
Note the history of the authorship of that specific book (and I'm not even talking about the _original_ origin as a fanfic) is a little less black and white. It was not "written on a BlackBerry". The author took notes on a BlackBerry while in transit. She then transferred those notes to her Mac and did the "writing" there. There's no statistic given on word count from the BB vs word count from the Mac. https://www.businessinsider.co... - I assume this is one of the articles the earlier poster read. I would wager real money that of the 156000-odd words in the final work, only a small proportion of them (maybe not even a single actual phrase) was typed on the BlackBerry; "notes" are not prose for publication.
BTW, remember how I said authors are not generally masochists? I think given the particular book we're talking about here, we can assume that this author doesn't fall into the general case.:P
Oh and by "fundamental change in the usage of language" I'm saying that a change of primary input method won't happen until language is not composed of words made of letters.
You've never written a book, I take it. (I have had three published, have written more, and am working on one in another window right now) Most authors are not masochists, at least not about writing. They use tools that are comfortable and efficient for the task ta hand, like any professional.
As a tour de force (or more likely as a sponsored demo) you could certainly write a book with a smartphone or tablet, but it's not efficient. The qwerty mechanical keyboard was developed specifically to address the problem of bulk text entry accessible to the majority of humankind at a good efficiency level. Touchscreens were designed for a totally different purpose, and text entry is an auxiliary capability of this input method.
Similarly, I could make a table using nothing but tools flaked from rocks. It wouldn't be a very good table, and I wouldn't be able to concentrate on the task of making it a better table, because I'd be too busy spending my energy struggling with inappropriate tools.
Let's be clear what we mean by "more". "More total words typed in a day"? Maybe (though I doubt it). But that is different from "more total words typed by a single person". A million people typing one line each on a smartphone is a totally different use case from a thousand people typing a thousand lines each on a keyboard.
Morse lasted so long because it was a training requirement for radio amateurs and the armed forces. Its advantage is that it can be transmitted over the barest possible shred of something that one can call a "communication medium" - if you have an on-state and an off-state, you can do Morse. It wasn't long-lived because of its efficiency, it was long-lived because it provided the best way to punch a message through an improvised or impaired communication medium.
Well, a) because of precisely what you just said - ASL isn't worldwide, whereas qwerty is (all standard keyboard layouts worldwide are very closely related to qwerty; azerty, qwertz, etc, including Asian and Cyrillic keyboards that have the Roman alphabet plus modifiers and additional IMs like guobi that give you characters from Roman typing), and b) because ASL wasn't designed to be used as an alphabet, it's a _language_ for which the letter signs are a fallback when there's no specific sign for the thing you need to say - so the letter signs aren't necessarily the easiest to remember, or "best" shapes from a muscle memory standpoint.
Mmm yes but this is hardly the full set of use cases for a keyboard. "people's data" being accessed through the cloud on a small screen is typically tiny amounts of data being accessed sporadically and far more reading than entry. The qwerty mechanical keyboard covers use cases such as entering and editing an entire program, writing a book, composing an email longer than a couple of paragraphs, and other things of that ilk. If we are saying that we want to remove the keyboard, then we probably have to make a fundamental change in the usage of language. Or we divide people even further into creators, who have full tools that can generate content, and consumers, who have crippled terminals that can only view content and provide short social media replies.
Input methods for a mobile terminal on which one enters small amounts of content in a burst reply to a short message have different problems to solve. For example, voice input is almost semi viable here (I personally hate it for many reasons, including privacy and accuracy), where it would be completely ludicrous for, say, programming.
For people who are making their living at keyboards today - it's pretty much simple to answer this: "No". For those people, and likely for the type of work we are doing until that work becomes irrelevant, there isn't a feasible replacement for the keyboard. Our workloads were designed around text, and there really isn't a better way to enter text than a keyboard that uses all our fingers. While you could argue that layout X is better than layout Y because ergonomics, the simple numeric and historical dominance of the QWERTY keyboard ensures that everyone is familiar with it and it has a pre-bias in everyone's muscle memory.
However it's not necessarily the case that future workloads will involve the direct entry of large amounts of text. Right now everything we do is fundamentally supported by a raft of text entry - sourcecode, documents, etc. It is conceivable that future workloads might involve the manipulation of some other way of abstracting the same concepts. As a totally artificial example, if tomorrow's programming language is designed such that the "sourcecode" is an array of 3D blocks, then it's easy to conceive that the IDE for such a language could be a VR or AR interface where you pick up and place those blocks with your fingres.
One might argue that the dominance of the QWERTY keyboard as "the input method" is already challenged by touchscreens - which don't even have keyboards on them all the time. But of course the real question is not "will something else replace QWERTY as the dominant input method" but really "will something else replace QWERTY as the dominant input FOR SIMILAR QUANTITIES OF TEXT" - which is a very different question. Touchscreens obviously are terrible for this. Let the flames begin.
The evolutionary part of this software (the piece that says "let's vary parameters and see what results we get, and learn what makes the overall result better") is sort of the easy bit. It's the flow simulation part of it ("how do we create a simulation of the thing being tested, that will accurately transfer to real life") that's hard. Your evolutionary algorithm might be wonderful, and there's _lots_ of FOSS in that field, but it's worthless if it's optimizing a simulation that doesn't accurately model real life. An illustrative article about this was published just the other day https://www.popularmechanics.c...
Fluid flow simulation is what one might call a military grade problem - efficient and accurate ways of doing it are either protected by commercial secretcy (because CAD software to design multimillion dollar yachts and aircraft is expensive) or actual military secrecy - because the problem you're solving is the same sort of problem that's being solved (for example) when designing SSBN propellers and hulls to minimize cavitation and make the ships run silent.
Heh. So, all the points you made there are very valid. I should have said more clearly what I meant, which is very tightly scoped: If you do not wish to be subjected to the particular surveillance activities that occur at airports, you have the option not to visit airports. Of course there is a large and growing surveillance web elsewhere also, though thankfully not so deep (or so well accepted) here in the US as in somewhere like, say, the UK. I would like to believe that there could be a revolt against it, but this seems unlikely.
In the case of drivers licenses and searches for alcohol, the reason for that phase in the paperwork is, of course, specifically to waive your Fourth Amendment rights such that no lawsuits or arguments can be raised. Driving without a license is a crime that carries a higher penalty than a first time DUI, so I wouldn't recommend it.
I think the reason they don't allow "useful" identification by facial recognition (and by "useful" here I mean "useful to you, so you don't have to carry papers") is because facial matches are probabilistic and they believe that requiring a physical token reduces the chance that someone will slip through with false identification. Also, facial recognition is fairly easy to spoof if you're determined (great example: https://creators.vice.com/en_u... are SUPER realistic). Note that for domestic travel, at least, it _is_ possible to travel and get through security checkpoints without ID, it's just amazingly irksome. One of my employees traveling with me left his wallet on an airplane and had no ID when flying back the next day, so I got a good description of the process; it's basically an Experian identity check asking you for various info out of your credit report.
Records of who enters and leaves the country are already kept, and in fact have been kept for decades. This isn't a new unreasonable search of any kind. So that side of the argument is irrelevant.
As for facial images - US (and other) passports already require expression-neutral images explicitly for the purpose of machine recognition to make it impossible for multiple passports to be issued to a single individual. If you have a passport issued in the last decade or so, your face is ALREADY in "the system". Again, this isn't an unreasonable search and it's not creating a new security hole unless the scanning systems being installed have new, implementation-specific holes; as far as "the database" goes, you're already in there. If you've got GlobalEntry like me, your fingerprints are in there too. And while you are in the airport, you have no expectation of privacy. Ever been in a US airport? There are signs everywhere indicating that all persons and property are subject to search. You have the option not to enter if you do not wish to be searched.
Heh. Thanks. I did warn you I was too drunk to look into it properly in my original reply - it's been a real, real long time since I needed to write any DOS code (though I used to make a very decent living out of it many years ago);)
Since I'm currently still in Puerto Rico and currently still well steeped in rum, nothing has changed since my earlier post.
Cool story, bro. You don't even need to go any further than looking at the comparative size of binaries compiled for x86 vs amd64 to know exactly what kind of expert you are. Plus, even taking what you said at face value (and it's partly true) - the RAM FOOTPRINT OF THE PROGRAM AS IT EXECUTES IS LARGER.
It's absolutely true that MS has a STRONG incentive to make a 32-bit version of Win10 simply to increase the compatibility footprint. They did the same sort of thing with Win95, remember - they INTENTIONALLY decided to make Win95 GDI be 16-bit SPECIFICALLY to support computers that had only 4MB RAM (yes 4MB - it was a different time...:). Only the incentive was even stronger with Win10 because of all the other things MS gets out of that - more computers running the Win Store (like anyone goes looking in that tumbleweed farm...) and MORE TELEMETRY AND ADVERTISING! YAY!
I wonder if that calculus has changed now, since Win10 is no longer "really" a free upgrade. (yeah, yeah, you can still DL and install and register the OS if you claim you have accessibility needs). I mean, supporting 32-bit editions costs them a lot of developer hours. Eventually, they will assuredly discontinue it - but obviously eventually isn't now. And equally obviously, there was a strong reason for them to make the effort when they first released Win10.
It's incorrect to say there's no downside to a 64-bit OS. The binaries are larger, inherently, because the operands are twice the width. This is significant for RAM-constrained systems, which is one of the reasons (for example) that netbooks with small RAM always shipped with 32-bit OSes even though Microsoft was already elbow-deep in 64 bit code in that era. Even MS-Office editions had recommended RAM minima for the 32 vs 64 bit editions that were different (and they also stated something to the effect of "you won't notice any functional difference between the two editions except that Excel can handle more rows and columns in the 64-bit edition" or something like that).
That's the general answer. There is also a very specific answer in the case of Windows: 64-bit editions of Windows cannot run Win16 apps. There are still (FML) significant chunks of Win16 code out there, which everyone can agree is a pain in the ass but it's still a reality for some verticals. There may be some other compatibility considerations, too - right now I'm too drunk to check, but DOS emulation is different between the 32 and 64 bit editions.
Now add the bit where you explain how it benefits you or me to have a $1000/month phone bill so that SHOULD we one day decide to hike through the Yukon-Koyukuk Census Area (0.03 persons per square mile), we can have stellar cellphone coverage.
Sane decisions have to be made on infrastructure investment. Are telecoms companies nests of evil and incompetence, staffed by tumbling assholes everyone would love to see set on fire? Indubitably. Is it unreasonable of them not to build a $50,000 gigabit fiber link to one remote log cabin in the middle of Alaska, and amortize that cost onto every iPhone user in Manhattan? No.
Some of the problems elucidated in the responses here are solvable, though of course with a cost per node penalty. For example, one could hypothesize extremely frequency- and protocol-agile nodes, capable of using a wide variety of communications methods inside the ISM bands (and, since we're talking about a wet dream here anyway) hey why not licensed bands as well? Imagine a multi-band, multi-protocol GSM/CDMA/HSPA/LTE/iDEN/WiMax/FRS/CB/amateur/WiFi/Zigbee/433MHz/infra-red/unicornfart transceiver that can discover and talk to everything around it by whatever means possible. Assume that electrical power is free and always available. Of course this doesn't make the network impossible for the government to disrupt, merely more expensive. Take the problem "How do we establish pervasive, non-goverment-interdicted internet in North Korea?" as being a good example of the extreme in the case we're trying to solve. All the DPRK has to do is have a few directional broadband sniffers in their police vehicles, and the strength of will to summarily shoot any citizen they believe is operating a transmitter of any kind. (For all I know, the DPRK already does this. I sure believe they *would* do it if they felt it necessary). Pretty soon you'll have the *citizens* looking for the one asshole in their apartment block running one of these nodes, so they can shut him down before the police do, in order to avoid collateral damage.
The bandwidth issue is a different problem, and I'd argue one that is irrelevant - if you consider the use case of this samizdat network. The tl;dr here is that providing bandwidth to support commercial enterprises like Google or Netflix isn't free. All those fibers and satellites and switches and the BOFHs needed to keep them working 24/7 cost real money, but the money is spent because it earns *more* money back [theoretically, LOL]. You could make an argument that if every person in the US was willing to spend say $5/month on this enterprise, they'd collectively have the financial backing to keep a fairly sophisticated network operational - but this isn't true, partly because of the physical limitations of these mesh networks that everyone else has discussed above, and partly simply due to your $5/month, spent on *your* node, not being able to contribute to any sort of economy of scale. You can't get a formula 1 car by getting together with 299 of your friends and buying 300 Yugos. But on the other hand... for the actual use case where this network shines, you don't need one. Google has huge bandwidth because its customers have huge, very realtime needs and if those needs aren't met, those customers will go elsewhere. But getting an answer in 0.0001sec to "how are babby formed?" without the government knowing isn't the use case for people who need to let the world see cellphone pictures of that government digging mass graves or publicly executing teenagers. They need to exfiltrate small amounts of data anonymously past specific barriers, and their needs are not very realtime. If it takes 10 minutes to get a single JPEG past the gatekeeper, that's fine.
The important points here are a) that the data doesn't need to stay in this global mesh for its whole life - in fact, it becomes most useful when it gets onto the "real" networks, and b) most enduser traffic is not of a nature that requires this protection. It's not necessary to bathe the world in a sea of connectivity, it's merely necessary to dig modest sized tunnels under all the border fences and assume that the data will eventually reach a friendly environment.
Trying to solve the general problem of "unlimited-bandwidth, uncensorable communications between any two points on the globe" is not a rational goal for this type of network, and generally not a useful goal anyway. For any given piece of data that's subject to censorship, it's rarely subject to censorship worldwide. Once you've helped it to escape from wherever it's restricted, it can spread on The Man's networks, because most of The Man doesn't care. If there is some piece of info
You're aware, I assume, that the equilibrium position for globalization is a truly level playing field where the standard of living in a rural third world village is the same as that in, say New York City? (This assumes zero transportation costs, or a work mode, such as telecommuting, that doesn't require transportation).
(I won't call it "math"). This is exactly analogous to saying "five million penguin farts are unleashed every year. there are 300 million people in the USA. Therefore, over a 30 year period, a family of four is directly responsible for two penguin farts."
Slashdot Mirror
1- provide an algorithmic experience. That's difficult. The "fuzzy logic" machine learning systems everyone else is using exist because writing strict algorithms to surface all the relevant data is hard.
2- provide a crappy experience from their own AI engine trained on limited data.
3- provide a ??? experience by buying someone else's training dataset.
This isn't even an Apple-specific question. In a world where we're talking about data privacy, in MANY fields (calendar/email/browser data is only the tip of the iceberg - consider autonomous driving data for example, including everything every autonomous car captures with every one of its sensors on every drive) - where is the tradeoff between "we want this thing to look smart" vs "we don't want to feed the beast"?
The inherent lack of clarity is bad enough, but it pales in comparison to the real problem, which is that are very few good actors on the other end of the wire, and an end user has no way to scrutinize them. By the time it is divulged that SiteX is illegally ignoring DNT and storing information outside whatever the local law permits, it's too late - the information is already sold to a thousand different data brokers and it is as undeletable as a nude selfie posted publicly to Facebook.
The only meaningful solution is to build the protections into the client side, so that the client is prevented from sending data that can be gathered by the server end. The server cannot, and can never be, trusted by the end-user. It's disappointing that the options we have in browsers (even with extensions) are still relatively coarse. For example, we need the ability to block all active scripting (including that embedded in a page, not just by blocking specific URLs to malware Javascript sources) except for a small whitelist of items critical to the function of the site. We need a way of blocking particular APIs from being accessed by active web content (there is NO reason why a website needs to know my battery level on a mobile device. If there was a reason, it would be a very limited use case that I would only enable for that one particular site. Same principle applies to a lot of the data that's used to fingerprint browsers).
For me, it comes down to "is this making code that runs more verifiably, or that can be developed faster", or some other advantage. I haven't yet seen a compelling argument - but I haven't seen proof that it's impossible either, so I continue to wait and see, and in the meantime continue to grovel in C and assembler because that's the fastest way for me to get little personal projects up and running.
Can there be other HMIs that work, and perhaps *in specific sets of constraints and for particular use cases* work better than qwerty? Sure. But that's not what I see when I read the original discussion. Swords at dawn, sir, on the exact semantics of the word "replace".
BTW, as you are now saying "how good could a mobile HMI be for making notes", that is a completely, completely different question from the original question, which was "can the qwerty keyboard be replaced?" (by implication superseded). A replacement for a thing has to exceed its capabilities in all use cases that matter. Yes, as I indicated, I too use a mobile tool where it's all I have available. But it's not a replacement for the real tool; it's just "the tool you have is better than no tool at all" and for me an electronic text entry tool with a loathsome entry method is slightly more useful than a moleskine because although I enjoy writing in a notebook much more than I enjoy poking about on a touchscreen, having my notes born digital so I can email them to myself and a) never lose them, and b) cut and paste directly into my WP.
In much the same way, a Swiss Army knife can be used to do a lot of emergency repairs or even surgery. And if it's all you've got, you use it. But you get better results, more consistently, and with a better user experience, when you're using a real set of tools or surgical equipment.
Note the history of the authorship of that specific book (and I'm not even talking about the _original_ origin as a fanfic) is a little less black and white. It was not "written on a BlackBerry". The author took notes on a BlackBerry while in transit. She then transferred those notes to her Mac and did the "writing" there. There's no statistic given on word count from the BB vs word count from the Mac. https://www.businessinsider.co... - I assume this is one of the articles the earlier poster read. I would wager real money that of the 156000-odd words in the final work, only a small proportion of them (maybe not even a single actual phrase) was typed on the BlackBerry; "notes" are not prose for publication.
BTW, remember how I said authors are not generally masochists? I think given the particular book we're talking about here, we can assume that this author doesn't fall into the general case. :P
Oh and by "fundamental change in the usage of language" I'm saying that a change of primary input method won't happen until language is not composed of words made of letters.
As a tour de force (or more likely as a sponsored demo) you could certainly write a book with a smartphone or tablet, but it's not efficient. The qwerty mechanical keyboard was developed specifically to address the problem of bulk text entry accessible to the majority of humankind at a good efficiency level. Touchscreens were designed for a totally different purpose, and text entry is an auxiliary capability of this input method.
Similarly, I could make a table using nothing but tools flaked from rocks. It wouldn't be a very good table, and I wouldn't be able to concentrate on the task of making it a better table, because I'd be too busy spending my energy struggling with inappropriate tools.
Let's be clear what we mean by "more". "More total words typed in a day"? Maybe (though I doubt it). But that is different from "more total words typed by a single person". A million people typing one line each on a smartphone is a totally different use case from a thousand people typing a thousand lines each on a keyboard.
Morse lasted so long because it was a training requirement for radio amateurs and the armed forces. Its advantage is that it can be transmitted over the barest possible shred of something that one can call a "communication medium" - if you have an on-state and an off-state, you can do Morse. It wasn't long-lived because of its efficiency, it was long-lived because it provided the best way to punch a message through an improvised or impaired communication medium.
Well, a) because of precisely what you just said - ASL isn't worldwide, whereas qwerty is (all standard keyboard layouts worldwide are very closely related to qwerty; azerty, qwertz, etc, including Asian and Cyrillic keyboards that have the Roman alphabet plus modifiers and additional IMs like guobi that give you characters from Roman typing), and b) because ASL wasn't designed to be used as an alphabet, it's a _language_ for which the letter signs are a fallback when there's no specific sign for the thing you need to say - so the letter signs aren't necessarily the easiest to remember, or "best" shapes from a muscle memory standpoint.
Input methods for a mobile terminal on which one enters small amounts of content in a burst reply to a short message have different problems to solve. For example, voice input is almost semi viable here (I personally hate it for many reasons, including privacy and accuracy), where it would be completely ludicrous for, say, programming.
However it's not necessarily the case that future workloads will involve the direct entry of large amounts of text. Right now everything we do is fundamentally supported by a raft of text entry - sourcecode, documents, etc. It is conceivable that future workloads might involve the manipulation of some other way of abstracting the same concepts. As a totally artificial example, if tomorrow's programming language is designed such that the "sourcecode" is an array of 3D blocks, then it's easy to conceive that the IDE for such a language could be a VR or AR interface where you pick up and place those blocks with your fingres.
One might argue that the dominance of the QWERTY keyboard as "the input method" is already challenged by touchscreens - which don't even have keyboards on them all the time. But of course the real question is not "will something else replace QWERTY as the dominant input method" but really "will something else replace QWERTY as the dominant input FOR SIMILAR QUANTITIES OF TEXT" - which is a very different question. Touchscreens obviously are terrible for this. Let the flames begin.
Fluid flow simulation is what one might call a military grade problem - efficient and accurate ways of doing it are either protected by commercial secretcy (because CAD software to design multimillion dollar yachts and aircraft is expensive) or actual military secrecy - because the problem you're solving is the same sort of problem that's being solved (for example) when designing SSBN propellers and hulls to minimize cavitation and make the ships run silent.
They can, through gasification. The Luftwaffe used this approach towards the end of the war.
In the case of drivers licenses and searches for alcohol, the reason for that phase in the paperwork is, of course, specifically to waive your Fourth Amendment rights such that no lawsuits or arguments can be raised. Driving without a license is a crime that carries a higher penalty than a first time DUI, so I wouldn't recommend it.
I think the reason they don't allow "useful" identification by facial recognition (and by "useful" here I mean "useful to you, so you don't have to carry papers") is because facial matches are probabilistic and they believe that requiring a physical token reduces the chance that someone will slip through with false identification. Also, facial recognition is fairly easy to spoof if you're determined (great example: https://creators.vice.com/en_u... are SUPER realistic). Note that for domestic travel, at least, it _is_ possible to travel and get through security checkpoints without ID, it's just amazingly irksome. One of my employees traveling with me left his wallet on an airplane and had no ID when flying back the next day, so I got a good description of the process; it's basically an Experian identity check asking you for various info out of your credit report.
As for facial images - US (and other) passports already require expression-neutral images explicitly for the purpose of machine recognition to make it impossible for multiple passports to be issued to a single individual. If you have a passport issued in the last decade or so, your face is ALREADY in "the system". Again, this isn't an unreasonable search and it's not creating a new security hole unless the scanning systems being installed have new, implementation-specific holes; as far as "the database" goes, you're already in there. If you've got GlobalEntry like me, your fingerprints are in there too. And while you are in the airport, you have no expectation of privacy. Ever been in a US airport? There are signs everywhere indicating that all persons and property are subject to search. You have the option not to enter if you do not wish to be searched.
Since I'm currently still in Puerto Rico and currently still well steeped in rum, nothing has changed since my earlier post.
Cool story, bro. You don't even need to go any further than looking at the comparative size of binaries compiled for x86 vs amd64 to know exactly what kind of expert you are. Plus, even taking what you said at face value (and it's partly true) - the RAM FOOTPRINT OF THE PROGRAM AS IT EXECUTES IS LARGER.
I wonder if that calculus has changed now, since Win10 is no longer "really" a free upgrade. (yeah, yeah, you can still DL and install and register the OS if you claim you have accessibility needs). I mean, supporting 32-bit editions costs them a lot of developer hours. Eventually, they will assuredly discontinue it - but obviously eventually isn't now. And equally obviously, there was a strong reason for them to make the effort when they first released Win10.
That's the general answer. There is also a very specific answer in the case of Windows: 64-bit editions of Windows cannot run Win16 apps. There are still (FML) significant chunks of Win16 code out there, which everyone can agree is a pain in the ass but it's still a reality for some verticals. There may be some other compatibility considerations, too - right now I'm too drunk to check, but DOS emulation is different between the 32 and 64 bit editions.
Sane decisions have to be made on infrastructure investment. Are telecoms companies nests of evil and incompetence, staffed by tumbling assholes everyone would love to see set on fire? Indubitably. Is it unreasonable of them not to build a $50,000 gigabit fiber link to one remote log cabin in the middle of Alaska, and amortize that cost onto every iPhone user in Manhattan? No.
The bandwidth issue is a different problem, and I'd argue one that is irrelevant - if you consider the use case of this samizdat network. The tl;dr here is that providing bandwidth to support commercial enterprises like Google or Netflix isn't free. All those fibers and satellites and switches and the BOFHs needed to keep them working 24/7 cost real money, but the money is spent because it earns *more* money back [theoretically, LOL]. You could make an argument that if every person in the US was willing to spend say $5/month on this enterprise, they'd collectively have the financial backing to keep a fairly sophisticated network operational - but this isn't true, partly because of the physical limitations of these mesh networks that everyone else has discussed above, and partly simply due to your $5/month, spent on *your* node, not being able to contribute to any sort of economy of scale. You can't get a formula 1 car by getting together with 299 of your friends and buying 300 Yugos. But on the other hand... for the actual use case where this network shines, you don't need one. Google has huge bandwidth because its customers have huge, very realtime needs and if those needs aren't met, those customers will go elsewhere. But getting an answer in 0.0001sec to "how are babby formed?" without the government knowing isn't the use case for people who need to let the world see cellphone pictures of that government digging mass graves or publicly executing teenagers. They need to exfiltrate small amounts of data anonymously past specific barriers, and their needs are not very realtime. If it takes 10 minutes to get a single JPEG past the gatekeeper, that's fine.
The important points here are a) that the data doesn't need to stay in this global mesh for its whole life - in fact, it becomes most useful when it gets onto the "real" networks, and b) most enduser traffic is not of a nature that requires this protection. It's not necessary to bathe the world in a sea of connectivity, it's merely necessary to dig modest sized tunnels under all the border fences and assume that the data will eventually reach a friendly environment.
Trying to solve the general problem of "unlimited-bandwidth, uncensorable communications between any two points on the globe" is not a rational goal for this type of network, and generally not a useful goal anyway. For any given piece of data that's subject to censorship, it's rarely subject to censorship worldwide. Once you've helped it to escape from wherever it's restricted, it can spread on The Man's networks, because most of The Man doesn't care. If there is some piece of info
You're aware, I assume, that the equilibrium position for globalization is a truly level playing field where the standard of living in a rural third world village is the same as that in, say New York City? (This assumes zero transportation costs, or a work mode, such as telecommuting, that doesn't require transportation).
(I won't call it "math"). This is exactly analogous to saying "five million penguin farts are unleashed every year. there are 300 million people in the USA. Therefore, over a 30 year period, a family of four is directly responsible for two penguin farts."