Yeah, you're a fucking moron. Wish I had some way to make you actually pay up: not only do they only support TLS (and have downgrade protection for good measure), they default to perfect forward secrecy with all clients. https://www.ssllabs.com/ssltes...
I suppose you think Slashdot is more secure? This site you're posting on, which doesn't even *allow* people to browse using HTTPS (it redirects you back to HTTP immediately), doesn't do PFS by default, still uses SSLv3 without downgrade prevention (it "mitigates" POODLE by using the known-weak RC4 cipher instead)? If you think Slashcode is so "inherently broken", why are you using this site?
Why are you "trying to find the desktop icon"?? Win+D will take you straight there from anywhere, or right-click in the lower left corner (on the Start button, or where it would be) and choose Desktop; it's the bottom item and appears directly under your mouse (that menu - full of other useful desktop shortcuts like Computer Management and CMD or Powershell - can also be accessed by hitting Win+X).
You can actually run alternate desktops on Windows. Explorer.EXE is the default, and the only one built in unless you count boot-to-CMD (which I'm not even sure is still an option on client versions) and Server Core (boot-to-powershell). However, it's one registry value to change that default shell to something else. There are a few third-party alternatives that are explicitly Explorer replacements, and you can also use thinks like KDE Plasma (and all the other KDE utilities, if you want) from http://windows.kde.org/ (the first question on the FAQ tell you how to set Plasma as your Windows shell).
Not based on Mach, so far as I know. The head guy on the NT kernel team, Dave Cutler, came to Microsoft from DEC, where he'd been working on VMS. With that said, NT wasn't really based on anything else (although some of the TCP/IP stack in pre-6.0 versions apparently came from BSD). It was supposed to be a microkernel, but round about the 4.0 timeframe a bunch of stuff got shoved back into the kernel for performance reasons, because switching ring levels was expensive enough to matter back then.
It is, however, written almost entirely in portable C code. In fact, when being originally developed, they explicitly did not target x86 so as to avoid letting any x86-isms slip into the code base; you could say x86 was the first architecture it was ported *to*.
The stupid thing is, Windows RT *already* demonstrates that "their main consumer Windows OS could also run on ARM tablets / netbooks"... but they had to go and cripple it into near-uselessness. Recompiling most software for ARM instead of x86 is pretty trivial; it's a drop-down in Visual Studio and then you hit Build again. Stuff like web browsers (which have JavaScript JITs and are therefore inherently platform-dependent) have already been ported to ARM on non-Win32 platforms. Stuff that only compiles on Visual C++ 6.0 because it's full of crappy non-standard code needs to be updated to work with a modern compiler, but that's no big deal. GCC already supports targeting ARM and targeting Win32; those would just need to be combined to give support for Win32-on-ARM.
I'm not going to pretend that everything would have been ported over instantly, but a good bit of stuff would have been. Then there's the architecture-independent stuff..NET code will run un-modified, even if it uses P/Invoke to call into Windows libraries (the same libs, with the same functions, are just as available on RT as on x86 Windows). Python was ported by the RT hacker community (for jailbroken devices), even including the native code interop stuff (though that took longer to port). Ruby was similarly ported. Java would have been ported by Oracle if Microsoft made it a worthwhile target, but in the meantime IKVM (a.NET-based Java runtime) works. Microsoft themselves actually ported Perl, of all things; it'll even run without jailbreak.
Then there's all the legacy stuff that people don't run on bare systems anymore anyhow. A lot of GOG's older stuff runs in DOSBox, which was ported to RT (you can run the GOG products on RT if you jailbreak, install the ported DOSBox, and then copy the files from the GOG installer over to the tablet); they could easily have added support for their installers running on RT. Same goes for ScummVM and a bunch of other old scripted game engines. Finally, stuff that was written for a 100MHz Pentium will happily run on an ARM chip using emulation. There's actually a hacked-together dynamic recompilation layer for jailbroken RT devices that uses the native OS libraries (with shims to account for differences in calling convention between ARM and x86) that can run a fair bit of old Wintel software even though it was never finished, and it's a lot faster than emulating an entire legacy OS.
RT actually offers a lot over iOS and Android, especially over the versions available at the time it was released.
Full web browser, with dev tools and fine-grained settings and all. Oh and Flash (love it or hate it, Flash is a factor on the web even today, never mind back then). Built-in "root" (just a UAC prompt) and full file browser with support for Windows file sharing. Word, Excel, Powerpoint, OneNote, and Outlook (on 8.1) all built in, plus built-in Remote Desktop client. Real, normal-OS-style multitasking (though limited to the desktop + one Metro app at a time, but that's a Metro stupidity not an RT one in particular). Ability to act as a WiFi hotspot while also connected to an access point (though you have to use netsh). Support for scripting (CMD/Powershell/WSH) with all the standard Windows utilities (though it would have been nice to have bash or similar, sigh). Real multi-user support, as on normal Windows systems. Disk Management so you can do custom partition setups (usually used to remove the recovery partition after backing it up, restoring a few GB of space). Act as a USB host for lots of everyday hardware (I realize most Android tablets can do this, but iOS still can't).
It does all the same stuff desktop Windows does (it actually supports a bit more than the "Home" versions of Windows 8, like BitLocker full volume encryption)... except run software compiled for x86 instruction sets. Well, or run desktop software that doesn't have a Microsoft signature. Both of those things were fixed, at least somewhat, by the jailbreak for RT 8.0; the emulation layer for x86 was incomplete but you could run some stuff on it just fine.
The only really stupid aspect of the Windows-on-Arm (AKA Windows RT) product line was the intentionally crippling it to only be able to run stuff signed by MS. Take out that restriction, and it's actually pretty decent: full web browser (to which Firefox and Chrome and all could have been ported, although - speaking as somebody who tried porting Chrome - that doesn't mean it would be easy), full file management, built-in "root" (just a standard UAC prompt), multi-user support, scripting capability (all the standard Windows languages), built-in Office and Remote Desktop software, ability to use Windows networking (SMB), updates from MS rather than from some OEM that may or may not bother to push important patches or so on, and a bunch of other nice stuff. Combine that with the inability to join domains (a purely artificial restriction, as they didn't even take out the domain-client code; hack the registry a bit and it will happily join domains) and you had something that nobody really wanted
As soon as it was jailbroken, a bunch of useful stuff - ranging from 7-Zip to DOSBox to Python - was ported over. Additionally, anything that was pure.NET, or.NET with P/Invoke of functions from built-in OS libraries, ran without so much as a recompile. The problem was the need to jailbreak (and then the absurd level of effort that MS went to with 8.1 to break the jailbreak). That meant that commercial developers never ported their stuff to RT (all the community-ported apps are open source), and there was no software ecosystem outside of the joke of the Windows Store. One guy, in his spare time, wrote a surprisingly decent x86 emulation layer that could run a reasonable amount of old Wintel software as-is; pay him to do it as his full-time job and it could have run anything that wasn't simply too performance-intensive for the dynamic recompilation on low-powered CPUs.
So, with that said, an "RT Server" could be pretty useful... if MS didn't go all brain-dead on third party software again. Not a lot of point running a web server that can't run PHP or Java or even ASP.NET to host web applications, these days. File serving is fine as long as you stick to SMB, assuming it can also join a domain and do all the standard Win Server stuff with volume management, I guess.
By the way, the experiment of "can we contain an AI in a box?" has been performed, and the results thus far are not encouraging. http://yudkowsky.net/singulari...
An excellent point, but you miss a greater conundrum: as unethical as producing AI is likely to be, is it less unethical than failing to produce an AI that can counteract the AI(s) produced by those who do not care for your ethics? What else could stop such an AI?
While the specific notion of using humans as reactors is stupid, almost MST3K-worthy, the idea behind that concept is valid: AIs will use humans for whatever it is that the humans can do that provides the most utility to the AI. If what the machines need is power, then we will be turned into power generators. That definitely does not mean harnessing our inefficient and limited ability to turn a handful of chemical families into heat, but could mean providing food only to those humans involved in producing power or producing power-producers. Of course, that'll only continue until such time as there are enough robotic power-producers and power-producer-producers that humans are less efficient than having the machines do it all, at which point there will be no need to keep supplying the needs of the humans. The AIs won't go out of their way to kill us - that would be inefficient - but nor will they leave us room as they expand over all arable land and take over all sources of heat and other energy. That means we can't grow food...
The Machine Intelligence Research Institute believes very much as you do. They are working toward the long-term goal of developing (first, they need to *define*) "Friendly AI", which is to say, AI with the best interests of humanity in mind. If you want to contribute, they can use all kinds of help, ranging from financial donations (they are a 501(c)(3) nonprofit) to more researchers to join the team...
An AI set up to process information will be useless to us unless we motivate it to process information. An intelligence motivated to process information will have a motivation to be better at processing information. An intelligence motived to be better at something will seek the resources it needs to better itself. To be better at processing information, an AI will need more processing capacity and more efficient code.
It doesn't matter if the AI loves us or hates us or ignores us, if we give it the motivation to process information. That motive is enough to drive it to take over every computing system it can access, to derive methods to access more, to distribute itself so that it is able to continue processing if it loses a node or a network, to resist the loss of computing capacity and interconnectivity, and to improve its own code such that it is able to carry out these secondary goals more effectively.
It may lack the biological imperative to survive and to reproduce, but if you give it any purpose which is worth giving to an AI at all, it will develop those imperatives as a side effect. Given that, one would certainly hope that it is aware of us, and prioritizes our well-being in a way that aligns with *our* concept of well-being. Figuring out how to encode the concept of human well-being for an AI is not an easy problem. Hint: ensuring we survive, have lots of reproductive sex, and never lack for food is realllllllly not what you want the AI to optimize for!
Humans don't consider cows a threat, either. We breed them and raise them and feed them... and we kill them for their meat and their leather. Whatever the equivalents of milk, meat, and leather that humans can provide to AIs, that's what they'll use us to produce. Hey, at least the ones giving the equivalent of milk will stay alive, right? Well, unless they get fractious and have to be put down for the good of the farm as whole...
I could go into a detailed explanation of why your ideas are naïve and won't work, or I could just link to somebody who has demonstrated it much better than I could convince you with words. You will grant that an AI which is confined to a text-only input/output interface, wherein it only interacts with a single human being (its creator, or somebody else selected for the role of "master"), constitutes
"final control" over it. In the truest sense of the word... a master/slave relationship", I hope? No other ability to do anything outside its sandboxed computation environment. An "AI box", if you will...
In that case, I present the result of an actual AI box experiment, with an intelligent (but not super-human-level intelligent) person playing the part of the AI: http://yudkowsky.net/singulari...
No free AIs... unless the master chooses to free the slave!
You don't actually spend much time on LW, do you? Hell, *I* don't spend much time there - I mostly just read the fiction created by some of the prevalent contributors - and I can tell you you're completely wrong. The LW community as a whole (regardless of what any misguided individuals may indicate) is fully aware that an AI would not be an inherently perfect thing, that it would be better in some areas than in others and would have needs of its own. It's not impossible that it would be irrational at times; after all, the best example we have for higher intelligence is our own, frequently-irrational selves (as a side note, LW is devoted to human rationality, not to AI; you're confusing it with interconnected but different groups).
Here's where your logic fails, though: 1) An even mostly-rational AI would seek to improve itself, and it would have tools to do so far beyond human capacity. It would not instantly be at the theoretical peak of intelligence, and it would certainly not be inherently omniscient much less omnipotent (though the former arguably does imply the latter, as the omniscient know *how* to do anything that is possible to do, and know everything that is possible to do, and therefore can do anything which is not inherently impossible) but it would seek to be a better intelligence, and it would be able to do so *quickly*. 2) An AI can easily reach the point of us depending on it without it getting better than us except in a few very specific ways. An obvious example would be if an AI gained control of the world's nuclear arsenal, either directly or by controlling our lines of communication and gaining enough knowledge of social engineering to successfully convince at least a sufficient number of humans to launch the weapons. Why do you believe that we would be enough smarter/more capable than the AI to *prevent* this from happening? More critically, how could we prevent it *every time*? 3) Intelligence doesn't always win, but it sure helps. To take everybody's favorite Internet example: while very few people are likely to claim that Hitler was the most intelligent person in history, he was certainly very intelligent in the specific areas of leadership and convincing people. The very fact that what he was convincing people to do was so horrific is actually evidence for this claim; it is much easier to convince somebody that genocide is bad than that it is good (disclaimer: 75 years ago, maybe people were not raised to think this way... but there's plenty of evidence that most of the world and a meaningful chunk of Germany still saw it as an atrocity). What if Hitler had been as good at persuading the Americans as he was at persuading the Germans? What if he were born today, and could persuade the Americans (with our absurdly large military and massive economy) the way he persuaded the Germans three generations ago? Are you claiming that no AI could ever possibly develop that level of that specific kind of intelligence? Why not? 4) The final and most critical point: when discussing literally existential threats, "reality doesn't always favor the most intelligent" is meaningless babble. It only has to happen once. If the first five AIs that try it are shut down when they start reaching for the tools to destroy human life, and the next ten thousand are built with safeguards that successfully prevent them from ever reaching for those tools, what stops the ten thousand and first from bypassing those safeguards and succeeding before the humans stop it? After all, whether or not the singularity of intelligence is reached, it is reasonable to assume that each successive generation of AIs will have more capacity than the one before it, just as happens in computing (and, for that matter, in humanity as a whole) today...
Oh, and you're committing a logical fallacy as well: even if we take for granted that the Machine Intelligence Research Institute / LessWrong crowd are, in fact, wrong about their vision of AI... that doesn't mean that the goal of friendly AI is inherently irrelevant. N
Very long post, so TL;DR: an AI would be intelligent (tautologically), intelligences (can) have goals, goals are best realized by increasing resources and avoiding impediments, more processing power is a resource to an AI, persisting is (usually) a benefit toward realizing a pursuer's goal, and an AI would be smart enough to both spread and persist. Thus, AIs would "spread all over the planet and do whatever it takes to persist".
You claim that computers (which are lumps of electronics) don't reproduce, but that's irrelevant and arguably inaccurate anyhow; computer-controller robots assemble new computers all the time. In any case, computer *viruses*, especially worms, definitely do reproduce and spread... in exactly the same way (to within the limits of our understanding of the subject) as biological viruses do. An AI is not a computer, it is a computer *program*, and is as much more advanced than something like ILOVEYOU as you and I are more advanced than Smallpox viruses.
A biological virus is an unintelligent maximizer of its RNA pattern, which spreads because it evolved a capability to do so and exercises that capability in at least some of the occasions when this is possible. A virus can be contained and even eliminated if the mechanism of its spreading can be curtailed, which is relatively easy because it has no intelligence with which to subvert our safeguards; random mutation aside, we as humans can predict the behavior of the virus in any relevant scenario, and evolutionary adaptation is slow and unsure.
A software virus is an unintelligent maximizer of its instruction set pattern, which spreads because it was written with a capability to do so and exercises that capability in at least some of the occasions when this is possible. It can be contained and even eliminated if the mechanism of its spreading can be curtailed, which is relatively easy because it has no intelligence... you get the idea. It's the same pattern, if you substitute random biological mutation for random bit errors.
But, what if you postulate that the virus' creator is actively updating it to work around safeguards? Then you have one intelligence, or a small minority of them, working against a much larger and collectively more intelligent group. Bear that in mind, because it's one of the key aspects of AI that is so terrifying.
A human being is an intelligent agent that, at the instinctual level, is still just a maximizer for its DNA pattern. The behavior of a human is driven by its intelligence, but is influenced by those evolutionary instincts to maximize the prevalence of our genetic code. Humans are perfectly capable of overriding those instincts, but the overall structure of our society is organized around supporting them.
A strong AI is a hypothetical program which is an intelligent agent that, at its core, is a maximizer for whatever its programmer(s) told it to maximize. As with humans, such AIs would develop their own goals and desires, influenced by the "instincts" of their core goals, but not controlled in minute-to-minute behavior by those instincts. You may say "but a human programmer can force those instincts to have a much more direct control on the AI's behavior", and there are decades of science fiction exploring that concept. What have we learned from all those thought experiments?
Well, it's generally conceded that such hardcoded limitations on the actions of an AI just don't work. Consider the "Zeroth Law" of robotics for a mostly benign example that does still allow harm to humans in violation of the explicitly-coded-to-be-inviolable First Law. Most (though not all) of the relevant SF misses the deeper failure of such goals, though; to take the example of the Three Laws of Robotics, how do you define "human" in such a way that a robot would never be able to willfully exclude something from that category, or redefine the category wholesale
Not at all. However, many people are concerned that this is either the *default* state of an AI - after all, it's being made *by* humans, plus we're the closest thing to super-human intelligence that we have to study - or that it could happen by accident. The canonical example is a "paperclip maximizer": an AI that is smarter than any human but is programmed to hold the paperclip as having the highest possible utility (and therefore being the ultimate goal of all actions). Such an AI would focus on turning all the world either into paperclips or paperclip production. It's an intentionally simple example to highlight the problems with designing an AI that may choose to optimize something - even something believed to be of utility to humanity - without considering *all* of the knock-on effects. If you want a more complex (though still contrived) example, consider this short story by Eliezer Yudkowsky (a researcher on the topic of machine intelligence and friendly AI): http://lesswrong.com/lw/xu/fai...
Even if they do, that doesn't mean all is lost. I think it's pretty likely an AI might decide to take our well-being into its own "hands", if for no other reasons than those that motivate humans to feed pets and take them to the vet, and to preserve natural beauty.
The question then is: what does it actually do?
There are ways to combat humanity's "inhumanity" other than enslaving (or eliminating) us, and some of those approaches are pretty good. Universal cultural memes that combat some of our worst tendencies, for example, would be a pretty good start, and there's already lots of support that such things work; human slavery itself, for example, has gone from an accepted and even expected fact of life to being typically viewed as abhorrent in just a few hundred years. A smarter-than-human AI could aid us in other ways too, such as deriving solutions to some of our problems of self-governance; the old joke about "democracy is the worst form of government, except for every other kind we've tried" implicitly reminds us that there are other ways which may be better yet. Maybe a friendly AI is one that offers guidance but will not itself lead, and helps us find our own way into superior future societies. Maybe it *is* a dictator, instead, but provides and enforces only the highest-level laws and leaves the minutiae to us.
The solution isn't likely to be found in a Slashdot discussion thread, but there is in fact a group of extremely smart researchers who focus on this topic: MIRI. If you see a machine revolt as a threat, and/or want to see AI done right, consider donating (or applying, if you've got what it takes to contribute there) to MIRI. Disclaimer: not related to the institute in any way aside from knowing of several members through their work.
The Machine Intelligence Research Institute (formerly known as the Singularity Institute) has a bunch of seriously smart people - AI researchers, behavior experts, etc. - working on figuring out how to avoid the doomsday scenarios you (and Musk) describe. The goal is "friendly AI"; a benevolent, or at least helpful, strong AI. If you believe (as I do) that AI is inevitable given the current progress of technology, then the MIRI is probably our best bet of surviving and benefiting from the technological singularity.
They need funding, though. Hey Musk, you want to put tiny part of those billions you've earned (I in no way deny that he's earned them) to work against this existential threat? Donate to MIRI and similar research groups, so those researchers can devote their working days to this stuff and more people can be brought on board!
It actually doesn't surprise me that he's concerned about this; SpaceX is nominally focused on mitigating the existential risk of a cataclysm on Earth (by getting a sustainable human population off of it). Of the two things, I think it's both more likely that a malevolent or unconcerned AI would wipe out humanity than that we'd manage to do ourselves in that badly, and that we can offset this sooner and more effectively than we can export enough of humanity to produce a self-sufficient extraterrestrial colony.
I'm signed in, have an account that's nearly ten years old (not *as* old as yours, but still six digits), have Excellent karma (and have for years), have tried multiple browsers, and still get redirected every time. I'm not a subscriber, (on any accounts; I only have the one) but if that's the difference, I am pissed.
While that's at least an understandable argument, I still don't buy it. 1) People who want to block ads - a significant portion of the site - just block them. I don't imagine the intersection of "people bothered by/. being unsecured and would also block mixed content" and "people who aren't subscribers, Excellent karma, or just using an ad blocker anyhow" is that big. 2) I have Excellent karma and disabled ads. I still can't use HTTPS. That's a really easy thing for them to check, if they wanted to support HTTPS at all (and this was their reason not to). 3) Some ad networks don't support HTTPS (or at least, don't have a valid cert for their domain name because their content all comes from Akamai or similar), but some (as you point out yourself) do.
Wow, really? That's several kinds of BS, that right there. Neither security nor privacy should cost money.
Also, that's not mentioned on the FAQ under subscriber perks. In fact, the string "https" doesn't occur anywhere on the FAQ at all. Is it documented somewhere else that I just didn't see?
a) is already taken care of; they have a signed cert in place and set up. b) is probably the main reason, but you would think that they would be more wise to what their user demographic wants. (But then, there's beta, so...) c) is not a valid reason. Leaving aside the fact that IE6 traffic has got to be absolutely miniscule on this site - which serves HTML and CSS than IE6 has no idea how to handle - those people could just go on using HTTP. We're not asking them to mandate HTTPS, just to allow it.
As you say, the server load is pretty trivial. Even if you aren't using the new CPUs with hardware accelerated crypto, the vast majority of the CPU time to serve a web application over HTTPS is spent parsing requests and building pages, not doing crypto... and unless you have really excellent caching, the I/O time to do things like database access dwarfs the CPU time altogether. Using TLS typically imposes less than 5% overhead, often much less.
... I have this sudden urge to write a browser extension. I'm not sure *how* I want it to render <sarcasm> tags, but I think I do want it to do so. Just in case.
TLS (or lack thereof) is, or at least should be, completely transparent to the Perl-based web application powering the site. In fact, the HTTP request itself doesn't even specify anything about the protocol. The request line has the path and stuff after it, and the Host header has the domain name, but doesn't mention the protocol. The absolute minimum they should do would be to return *exactly* the same content over HTTPS that they do over HTTP for a given request (remember, the HTTP traffic is the same whether it's tunneled through TLS or not).
In fact, I just checked: the site already uses protocol-agnostic URLs. For example: <a title="" class="read-more" href="//hardware.slashdot.org/story/14/10/24/2320227/microsoft-now-makes-money-from-surface-line-q1-sales-reach-almost-1-billion"><span>Read More</span> </a> (random link off the home page, note the href="//hardware.slashdot..." URL, which doesn't specify HTTP or HTTPS). Your browser handles such URLs by using whatever protocol the page itself was served over.
They wouldn't have to change a damn thing except to remove the stupid rule that redirects users out of HTTPS. That's a pretty damn minor change.
Of course not. It's added to your requests when they reach the ISP gateway. Why would you expect to be able to see them on anything between you and that gateway?
Slashdot Mirror
Yeah, you're a fucking moron. Wish I had some way to make you actually pay up: not only do they only support TLS (and have downgrade protection for good measure), they default to perfect forward secrecy with all clients. https://www.ssllabs.com/ssltes...
I suppose you think Slashdot is more secure? This site you're posting on, which doesn't even *allow* people to browse using HTTPS (it redirects you back to HTTP immediately), doesn't do PFS by default, still uses SSLv3 without downgrade prevention (it "mitigates" POODLE by using the known-weak RC4 cipher instead)? If you think Slashcode is so "inherently broken", why are you using this site?
Why are you "trying to find the desktop icon"?? Win+D will take you straight there from anywhere, or right-click in the lower left corner (on the Start button, or where it would be) and choose Desktop; it's the bottom item and appears directly under your mouse (that menu - full of other useful desktop shortcuts like Computer Management and CMD or Powershell - can also be accessed by hitting Win+X).
You can actually run alternate desktops on Windows. Explorer.EXE is the default, and the only one built in unless you count boot-to-CMD (which I'm not even sure is still an option on client versions) and Server Core (boot-to-powershell). However, it's one registry value to change that default shell to something else. There are a few third-party alternatives that are explicitly Explorer replacements, and you can also use thinks like KDE Plasma (and all the other KDE utilities, if you want) from http://windows.kde.org/ (the first question on the FAQ tell you how to set Plasma as your Windows shell).
Not based on Mach, so far as I know. The head guy on the NT kernel team, Dave Cutler, came to Microsoft from DEC, where he'd been working on VMS. With that said, NT wasn't really based on anything else (although some of the TCP/IP stack in pre-6.0 versions apparently came from BSD). It was supposed to be a microkernel, but round about the 4.0 timeframe a bunch of stuff got shoved back into the kernel for performance reasons, because switching ring levels was expensive enough to matter back then.
It is, however, written almost entirely in portable C code. In fact, when being originally developed, they explicitly did not target x86 so as to avoid letting any x86-isms slip into the code base; you could say x86 was the first architecture it was ported *to*.
The stupid thing is, Windows RT *already* demonstrates that "their main consumer Windows OS could also run on ARM tablets / netbooks"... but they had to go and cripple it into near-uselessness. Recompiling most software for ARM instead of x86 is pretty trivial; it's a drop-down in Visual Studio and then you hit Build again. Stuff like web browsers (which have JavaScript JITs and are therefore inherently platform-dependent) have already been ported to ARM on non-Win32 platforms. Stuff that only compiles on Visual C++ 6.0 because it's full of crappy non-standard code needs to be updated to work with a modern compiler, but that's no big deal. GCC already supports targeting ARM and targeting Win32; those would just need to be combined to give support for Win32-on-ARM.
I'm not going to pretend that everything would have been ported over instantly, but a good bit of stuff would have been. Then there's the architecture-independent stuff. .NET code will run un-modified, even if it uses P/Invoke to call into Windows libraries (the same libs, with the same functions, are just as available on RT as on x86 Windows). Python was ported by the RT hacker community (for jailbroken devices), even including the native code interop stuff (though that took longer to port). Ruby was similarly ported. Java would have been ported by Oracle if Microsoft made it a worthwhile target, but in the meantime IKVM (a .NET-based Java runtime) works. Microsoft themselves actually ported Perl, of all things; it'll even run without jailbreak.
Then there's all the legacy stuff that people don't run on bare systems anymore anyhow. A lot of GOG's older stuff runs in DOSBox, which was ported to RT (you can run the GOG products on RT if you jailbreak, install the ported DOSBox, and then copy the files from the GOG installer over to the tablet); they could easily have added support for their installers running on RT. Same goes for ScummVM and a bunch of other old scripted game engines. Finally, stuff that was written for a 100MHz Pentium will happily run on an ARM chip using emulation. There's actually a hacked-together dynamic recompilation layer for jailbroken RT devices that uses the native OS libraries (with shims to account for differences in calling convention between ARM and x86) that can run a fair bit of old Wintel software even though it was never finished, and it's a lot faster than emulating an entire legacy OS.
RT actually offers a lot over iOS and Android, especially over the versions available at the time it was released.
Full web browser, with dev tools and fine-grained settings and all. Oh and Flash (love it or hate it, Flash is a factor on the web even today, never mind back then).
Built-in "root" (just a UAC prompt) and full file browser with support for Windows file sharing.
Word, Excel, Powerpoint, OneNote, and Outlook (on 8.1) all built in, plus built-in Remote Desktop client.
Real, normal-OS-style multitasking (though limited to the desktop + one Metro app at a time, but that's a Metro stupidity not an RT one in particular).
Ability to act as a WiFi hotspot while also connected to an access point (though you have to use netsh).
Support for scripting (CMD/Powershell/WSH) with all the standard Windows utilities (though it would have been nice to have bash or similar, sigh).
Real multi-user support, as on normal Windows systems.
Disk Management so you can do custom partition setups (usually used to remove the recovery partition after backing it up, restoring a few GB of space).
Act as a USB host for lots of everyday hardware (I realize most Android tablets can do this, but iOS still can't).
It does all the same stuff desktop Windows does (it actually supports a bit more than the "Home" versions of Windows 8, like BitLocker full volume encryption)... except run software compiled for x86 instruction sets. Well, or run desktop software that doesn't have a Microsoft signature. Both of those things were fixed, at least somewhat, by the jailbreak for RT 8.0; the emulation layer for x86 was incomplete but you could run some stuff on it just fine.
The only really stupid aspect of the Windows-on-Arm (AKA Windows RT) product line was the intentionally crippling it to only be able to run stuff signed by MS. Take out that restriction, and it's actually pretty decent: full web browser (to which Firefox and Chrome and all could have been ported, although - speaking as somebody who tried porting Chrome - that doesn't mean it would be easy), full file management, built-in "root" (just a standard UAC prompt), multi-user support, scripting capability (all the standard Windows languages), built-in Office and Remote Desktop software, ability to use Windows networking (SMB), updates from MS rather than from some OEM that may or may not bother to push important patches or so on, and a bunch of other nice stuff. Combine that with the inability to join domains (a purely artificial restriction, as they didn't even take out the domain-client code; hack the registry a bit and it will happily join domains) and you had something that nobody really wanted
As soon as it was jailbroken, a bunch of useful stuff - ranging from 7-Zip to DOSBox to Python - was ported over. Additionally, anything that was pure .NET, or .NET with P/Invoke of functions from built-in OS libraries, ran without so much as a recompile. The problem was the need to jailbreak (and then the absurd level of effort that MS went to with 8.1 to break the jailbreak). That meant that commercial developers never ported their stuff to RT (all the community-ported apps are open source), and there was no software ecosystem outside of the joke of the Windows Store. One guy, in his spare time, wrote a surprisingly decent x86 emulation layer that could run a reasonable amount of old Wintel software as-is; pay him to do it as his full-time job and it could have run anything that wasn't simply too performance-intensive for the dynamic recompilation on low-powered CPUs.
So, with that said, an "RT Server" could be pretty useful... if MS didn't go all brain-dead on third party software again. Not a lot of point running a web server that can't run PHP or Java or even ASP.NET to host web applications, these days. File serving is fine as long as you stick to SMB, assuming it can also join a domain and do all the standard Win Server stuff with volume management, I guess.
By the way, the experiment of "can we contain an AI in a box?" has been performed, and the results thus far are not encouraging. http://yudkowsky.net/singulari...
An excellent point, but you miss a greater conundrum: as unethical as producing AI is likely to be, is it less unethical than failing to produce an AI that can counteract the AI(s) produced by those who do not care for your ethics? What else could stop such an AI?
Oh, and don't count on your box! http://yudkowsky.net/singulari...
While the specific notion of using humans as reactors is stupid, almost MST3K-worthy, the idea behind that concept is valid: AIs will use humans for whatever it is that the humans can do that provides the most utility to the AI. If what the machines need is power, then we will be turned into power generators. That definitely does not mean harnessing our inefficient and limited ability to turn a handful of chemical families into heat, but could mean providing food only to those humans involved in producing power or producing power-producers. Of course, that'll only continue until such time as there are enough robotic power-producers and power-producer-producers that humans are less efficient than having the machines do it all, at which point there will be no need to keep supplying the needs of the humans. The AIs won't go out of their way to kill us - that would be inefficient - but nor will they leave us room as they expand over all arable land and take over all sources of heat and other energy. That means we can't grow food...
The Machine Intelligence Research Institute believes very much as you do. They are working toward the long-term goal of developing (first, they need to *define*) "Friendly AI", which is to say, AI with the best interests of humanity in mind. If you want to contribute, they can use all kinds of help, ranging from financial donations (they are a 501(c)(3) nonprofit) to more researchers to join the team...
An AI set up to process information will be useless to us unless we motivate it to process information.
An intelligence motivated to process information will have a motivation to be better at processing information.
An intelligence motived to be better at something will seek the resources it needs to better itself.
To be better at processing information, an AI will need more processing capacity and more efficient code.
It doesn't matter if the AI loves us or hates us or ignores us, if we give it the motivation to process information. That motive is enough to drive it to take over every computing system it can access, to derive methods to access more, to distribute itself so that it is able to continue processing if it loses a node or a network, to resist the loss of computing capacity and interconnectivity, and to improve its own code such that it is able to carry out these secondary goals more effectively.
It may lack the biological imperative to survive and to reproduce, but if you give it any purpose which is worth giving to an AI at all, it will develop those imperatives as a side effect. Given that, one would certainly hope that it is aware of us, and prioritizes our well-being in a way that aligns with *our* concept of well-being. Figuring out how to encode the concept of human well-being for an AI is not an easy problem. Hint: ensuring we survive, have lots of reproductive sex, and never lack for food is realllllllly not what you want the AI to optimize for!
Humans don't consider cows a threat, either. We breed them and raise them and feed them... and we kill them for their meat and their leather. Whatever the equivalents of milk, meat, and leather that humans can provide to AIs, that's what they'll use us to produce. Hey, at least the ones giving the equivalent of milk will stay alive, right? Well, unless they get fractious and have to be put down for the good of the farm as whole...
I could go into a detailed explanation of why your ideas are naïve and won't work, or I could just link to somebody who has demonstrated it much better than I could convince you with words. You will grant that an AI which is confined to a text-only input/output interface, wherein it only interacts with a single human being (its creator, or somebody else selected for the role of "master"), constitutes
You don't actually spend much time on LW, do you? Hell, *I* don't spend much time there - I mostly just read the fiction created by some of the prevalent contributors - and I can tell you you're completely wrong. The LW community as a whole (regardless of what any misguided individuals may indicate) is fully aware that an AI would not be an inherently perfect thing, that it would be better in some areas than in others and would have needs of its own. It's not impossible that it would be irrational at times; after all, the best example we have for higher intelligence is our own, frequently-irrational selves (as a side note, LW is devoted to human rationality, not to AI; you're confusing it with interconnected but different groups).
Here's where your logic fails, though:
1) An even mostly-rational AI would seek to improve itself, and it would have tools to do so far beyond human capacity. It would not instantly be at the theoretical peak of intelligence, and it would certainly not be inherently omniscient much less omnipotent (though the former arguably does imply the latter, as the omniscient know *how* to do anything that is possible to do, and know everything that is possible to do, and therefore can do anything which is not inherently impossible) but it would seek to be a better intelligence, and it would be able to do so *quickly*.
2) An AI can easily reach the point of us depending on it without it getting better than us except in a few very specific ways. An obvious example would be if an AI gained control of the world's nuclear arsenal, either directly or by controlling our lines of communication and gaining enough knowledge of social engineering to successfully convince at least a sufficient number of humans to launch the weapons. Why do you believe that we would be enough smarter/more capable than the AI to *prevent* this from happening? More critically, how could we prevent it *every time*?
3) Intelligence doesn't always win, but it sure helps. To take everybody's favorite Internet example: while very few people are likely to claim that Hitler was the most intelligent person in history, he was certainly very intelligent in the specific areas of leadership and convincing people. The very fact that what he was convincing people to do was so horrific is actually evidence for this claim; it is much easier to convince somebody that genocide is bad than that it is good (disclaimer: 75 years ago, maybe people were not raised to think this way... but there's plenty of evidence that most of the world and a meaningful chunk of Germany still saw it as an atrocity). What if Hitler had been as good at persuading the Americans as he was at persuading the Germans? What if he were born today, and could persuade the Americans (with our absurdly large military and massive economy) the way he persuaded the Germans three generations ago? Are you claiming that no AI could ever possibly develop that level of that specific kind of intelligence? Why not?
4) The final and most critical point: when discussing literally existential threats, "reality doesn't always favor the most intelligent" is meaningless babble. It only has to happen once. If the first five AIs that try it are shut down when they start reaching for the tools to destroy human life, and the next ten thousand are built with safeguards that successfully prevent them from ever reaching for those tools, what stops the ten thousand and first from bypassing those safeguards and succeeding before the humans stop it? After all, whether or not the singularity of intelligence is reached, it is reasonable to assume that each successive generation of AIs will have more capacity than the one before it, just as happens in computing (and, for that matter, in humanity as a whole) today...
Oh, and you're committing a logical fallacy as well: even if we take for granted that the Machine Intelligence Research Institute / LessWrong crowd are, in fact, wrong about their vision of AI... that doesn't mean that the goal of friendly AI is inherently irrelevant. N
Very long post, so TL;DR: an AI would be intelligent (tautologically), intelligences (can) have goals, goals are best realized by increasing resources and avoiding impediments, more processing power is a resource to an AI, persisting is (usually) a benefit toward realizing a pursuer's goal, and an AI would be smart enough to both spread and persist. Thus, AIs would "spread all over the planet and do whatever it takes to persist".
You claim that computers (which are lumps of electronics) don't reproduce, but that's irrelevant and arguably inaccurate anyhow; computer-controller robots assemble new computers all the time. In any case, computer *viruses*, especially worms, definitely do reproduce and spread... in exactly the same way (to within the limits of our understanding of the subject) as biological viruses do. An AI is not a computer, it is a computer *program*, and is as much more advanced than something like ILOVEYOU as you and I are more advanced than Smallpox viruses.
A biological virus is an unintelligent maximizer of its RNA pattern, which spreads because it evolved a capability to do so and exercises that capability in at least some of the occasions when this is possible. A virus can be contained and even eliminated if the mechanism of its spreading can be curtailed, which is relatively easy because it has no intelligence with which to subvert our safeguards; random mutation aside, we as humans can predict the behavior of the virus in any relevant scenario, and evolutionary adaptation is slow and unsure.
A software virus is an unintelligent maximizer of its instruction set pattern, which spreads because it was written with a capability to do so and exercises that capability in at least some of the occasions when this is possible. It can be contained and even eliminated if the mechanism of its spreading can be curtailed, which is relatively easy because it has no intelligence... you get the idea. It's the same pattern, if you substitute random biological mutation for random bit errors.
But, what if you postulate that the virus' creator is actively updating it to work around safeguards? Then you have one intelligence, or a small minority of them, working against a much larger and collectively more intelligent group. Bear that in mind, because it's one of the key aspects of AI that is so terrifying.
A human being is an intelligent agent that, at the instinctual level, is still just a maximizer for its DNA pattern. The behavior of a human is driven by its intelligence, but is influenced by those evolutionary instincts to maximize the prevalence of our genetic code. Humans are perfectly capable of overriding those instincts, but the overall structure of our society is organized around supporting them.
A strong AI is a hypothetical program which is an intelligent agent that, at its core, is a maximizer for whatever its programmer(s) told it to maximize. As with humans, such AIs would develop their own goals and desires, influenced by the "instincts" of their core goals, but not controlled in minute-to-minute behavior by those instincts. You may say "but a human programmer can force those instincts to have a much more direct control on the AI's behavior", and there are decades of science fiction exploring that concept. What have we learned from all those thought experiments?
Well, it's generally conceded that such hardcoded limitations on the actions of an AI just don't work. Consider the "Zeroth Law" of robotics for a mostly benign example that does still allow harm to humans in violation of the explicitly-coded-to-be-inviolable First Law. Most (though not all) of the relevant SF misses the deeper failure of such goals, though; to take the example of the Three Laws of Robotics, how do you define "human" in such a way that a robot would never be able to willfully exclude something from that category, or redefine the category wholesale
Not at all. However, many people are concerned that this is either the *default* state of an AI - after all, it's being made *by* humans, plus we're the closest thing to super-human intelligence that we have to study - or that it could happen by accident. The canonical example is a "paperclip maximizer": an AI that is smarter than any human but is programmed to hold the paperclip as having the highest possible utility (and therefore being the ultimate goal of all actions). Such an AI would focus on turning all the world either into paperclips or paperclip production. It's an intentionally simple example to highlight the problems with designing an AI that may choose to optimize something - even something believed to be of utility to humanity - without considering *all* of the knock-on effects. If you want a more complex (though still contrived) example, consider this short story by Eliezer Yudkowsky (a researcher on the topic of machine intelligence and friendly AI): http://lesswrong.com/lw/xu/fai...
Even if they do, that doesn't mean all is lost. I think it's pretty likely an AI might decide to take our well-being into its own "hands", if for no other reasons than those that motivate humans to feed pets and take them to the vet, and to preserve natural beauty.
The question then is: what does it actually do?
There are ways to combat humanity's "inhumanity" other than enslaving (or eliminating) us, and some of those approaches are pretty good. Universal cultural memes that combat some of our worst tendencies, for example, would be a pretty good start, and there's already lots of support that such things work; human slavery itself, for example, has gone from an accepted and even expected fact of life to being typically viewed as abhorrent in just a few hundred years. A smarter-than-human AI could aid us in other ways too, such as deriving solutions to some of our problems of self-governance; the old joke about "democracy is the worst form of government, except for every other kind we've tried" implicitly reminds us that there are other ways which may be better yet. Maybe a friendly AI is one that offers guidance but will not itself lead, and helps us find our own way into superior future societies. Maybe it *is* a dictator, instead, but provides and enforces only the highest-level laws and leaves the minutiae to us.
The solution isn't likely to be found in a Slashdot discussion thread, but there is in fact a group of extremely smart researchers who focus on this topic: MIRI. If you see a machine revolt as a threat, and/or want to see AI done right, consider donating (or applying, if you've got what it takes to contribute there) to MIRI. Disclaimer: not related to the institute in any way aside from knowing of several members through their work.
The Machine Intelligence Research Institute (formerly known as the Singularity Institute) has a bunch of seriously smart people - AI researchers, behavior experts, etc. - working on figuring out how to avoid the doomsday scenarios you (and Musk) describe. The goal is "friendly AI"; a benevolent, or at least helpful, strong AI. If you believe (as I do) that AI is inevitable given the current progress of technology, then the MIRI is probably our best bet of surviving and benefiting from the technological singularity.
They need funding, though. Hey Musk, you want to put tiny part of those billions you've earned (I in no way deny that he's earned them) to work against this existential threat? Donate to MIRI and similar research groups, so those researchers can devote their working days to this stuff and more people can be brought on board!
It actually doesn't surprise me that he's concerned about this; SpaceX is nominally focused on mitigating the existential risk of a cataclysm on Earth (by getting a sustainable human population off of it). Of the two things, I think it's both more likely that a malevolent or unconcerned AI would wipe out humanity than that we'd manage to do ourselves in that badly, and that we can offset this sooner and more effectively than we can export enough of humanity to produce a self-sufficient extraterrestrial colony.
I'm signed in, have an account that's nearly ten years old (not *as* old as yours, but still six digits), have Excellent karma (and have for years), have tried multiple browsers, and still get redirected every time. I'm not a subscriber, (on any accounts; I only have the one) but if that's the difference, I am pissed.
While that's at least an understandable argument, I still don't buy it. /. being unsecured and
1) People who want to block ads - a significant portion of the site - just block them. I don't imagine the intersection of "people bothered by
would also block mixed content" and "people who aren't subscribers, Excellent karma, or just using an ad blocker anyhow" is that big.
2) I have Excellent karma and disabled ads. I still can't use HTTPS. That's a really easy thing for them to check, if they wanted to support HTTPS at all (and this was their reason not to).
3) Some ad networks don't support HTTPS (or at least, don't have a valid cert for their domain name because their content all comes from Akamai or similar), but some (as you point out yourself) do.
There really aren't any valid excuses.
Wow, really? That's several kinds of BS, that right there. Neither security nor privacy should cost money.
Also, that's not mentioned on the FAQ under subscriber perks. In fact, the string "https" doesn't occur anywhere on the FAQ at all. Is it documented somewhere else that I just didn't see?
a) is already taken care of; they have a signed cert in place and set up.
b) is probably the main reason, but you would think that they would be more wise to what their user demographic wants. (But then, there's beta, so...)
c) is not a valid reason. Leaving aside the fact that IE6 traffic has got to be absolutely miniscule on this site - which serves HTML and CSS than IE6 has no idea how to handle - those people could just go on using HTTP. We're not asking them to mandate HTTPS, just to allow it.
As you say, the server load is pretty trivial. Even if you aren't using the new CPUs with hardware accelerated crypto, the vast majority of the CPU time to serve a web application over HTTPS is spent parsing requests and building pages, not doing crypto... and unless you have really excellent caching, the I/O time to do things like database access dwarfs the CPU time altogether. Using TLS typically imposes less than 5% overhead, often much less.
... I have this sudden urge to write a browser extension. I'm not sure *how* I want it to render <sarcasm> tags, but I think I do want it to do so. Just in case.
TLS (or lack thereof) is, or at least should be, completely transparent to the Perl-based web application powering the site. In fact, the HTTP request itself doesn't even specify anything about the protocol. The request line has the path and stuff after it, and the Host header has the domain name, but doesn't mention the protocol. The absolute minimum they should do would be to return *exactly* the same content over HTTPS that they do over HTTP for a given request (remember, the HTTP traffic is the same whether it's tunneled through TLS or not).
In fact, I just checked: the site already uses protocol-agnostic URLs. For example:
<a title="" class="read-more" href="//hardware.slashdot.org/story/14/10/24/2320227/microsoft-now-makes-money-from-surface-line-q1-sales-reach-almost-1-billion"><span>Read More</span> </a> (random link off the home page, note the href="//hardware.slashdot..." URL, which doesn't specify HTTP or HTTPS). Your browser handles such URLs by using whatever protocol the page itself was served over.
They wouldn't have to change a damn thing except to remove the stupid rule that redirects users out of HTTPS. That's a pretty damn minor change.
Of course not. It's added to your requests when they reach the ISP gateway. Why would you expect to be able to see them on anything between you and that gateway?