Wikipedia:Reference desk/Archives/Computing/2011 October 5

Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

October 5

What could cause a download to 'shrink' in size?

Partly Resolved. Second try was successful 220.101 talk^\Contribs 02:00, 7 October 2011 (UTC)

Recently I downloaded a "SystemRescueCD" (systemrescuecd-x86-2.3.1.iso) from "www.sysresccd.org/". It took several hours and about 352 Mb of bandwidth measured using a utility called NetWorx . However, when it finished saving to my HDD as a Disc Image File it was only "14.1 MB (14,876,672 bytes)". Has my download failed (seems likely) or is there another explanation for this? 220.101 talk^\Contribs(aka user:220.101.28.25) 02:09, 5 October 2011 (UTC)

md5sum systemrescuecd-x86-2.3.1.iso — if it doesn't spit out 8813aa38506f6e6be1c02e871eb898ca, then the image is no good. ¦ Reisio (talk) 11:34, 5 October 2011 (UTC)

Thanks for reply Reisio. Does that check have to be run via a command line? (nb. I am using WinXP, if that makes any diff.) I added a title to the question too!- 220.101 talk^\Contribs 17:59, 5 October 2011 (UTC)

Yes. You can get graphical md5/sha1/etc. sum checking apps, but that's out of my jurisdiction. ¦ Reisio (talk) 19:07, 5 October 2011 (UTC)

With 14.1MB, it can't be right. Alternative link: http://en.sourceforge.jp/.. systemrescuecd-x86-2.3.1.iso/ (it's also available as torrent btw) DS Belgium (talk) 23:08, 5 October 2011 (UTC)

Something went wrong. The download page says the file should be 352 megabytes. Or more specifically, 352 MiB (mebibytes, binary megabytes, 2²⁰ bytes). If I use wget to see the response headers for the download link, it says the file is exactly 369,342,464 bytes. --Bavi H (talk) 01:50, 6 October 2011 (UTC)

I agree Bavi H, and DS Belgium, something is wrong!. But the 'Date modified' (Monday, 3 October 2011, 1:02:12 PM) was when I started the download and 'Date created' when it finished at 4:16:34 PM. I was downloading something, at up to 110 Mb per hour. Maybe I saved it somewhere else? (Nope, did a search of my HDD, no luck!) - 220.101 talk^\Contribs 12:11, 6 October 2011 (UTC)

  Downloaded again, (3 hours and 50 minutes!), 352 MB (369,342,464 bytes). Seems it worked, though still wondering what happened the first time!
Thanks to all those who tried to help. - 220.101 talk^\Contribs 02:00, 7 October 2011 (UTC)

If an A.I.'s objective is to improve itself, couldn't that speed up and spiral out of control?

Say a new supercomputer that's not built yet will have the objective to find quicker, easier and lower-cost ways to clean up the planet and bring harmony/eudaimonia to the whole human race. While on the quest to find these solutions, it's also given a secondary objective to improve itself (its algorithms, processes, hardware composition, et al.) so that it can teach/equip itself to heal humanity even faster.

Once it starts on its secondary objective, wouldn't it then gain the ability to work faster, even on its own secondary objective? Therefore, when it improves itself even faster, it not only accelerates its own self-improvement, but it accelerates the acceleration of its own self-improvement. This would be something of a recursive feedback loop.

What happens if said loop gets out of control? What is that phenomenon called, and what else will come out of this when this happens? How do we keep control of this phenomenon? --70.179.161.22 (talk) 11:55, 5 October 2011 (UTC)

We make sure we build in the Three Laws of Robotics. Mitch Ames (talk) 12:40, 5 October 2011 (UTC)

And an off switch! ;) - 220.101 talk^\Contribs 13:11, 5 October 2011 (UTC)

I think it is a legitimate concern. As far as I know nobody has ever designed a system that works that way -- it would be like having a computer that is capable of swapping boards inside its box and editing its own boot ROM: the danger of instability would be pretty high. I'm not aware of a name for that sort of instability, though -- I expect Douglas Hofstadter would call it a kindo of strange loop. Looie496 (talk) 13:46, 5 October 2011 (UTC)

See technological singularity. --Goodbye Galaxy (talk) 14:36, 5 October 2011 (UTC)

It's hard to imagine a singularity happening. After decades of research, we have not come to a good understanding of how intelligence works, and so we don't have anything like a roadmap for producing a generally intelligent program. If, after a couple of centuries, we were able to produce a program as smart as a human, it wouldn't contribute any more to the research effort than raising a kid who grows up to be an AI researcher. We would need to produce something not as smart as a human genius, but far superior to one in order for this to happen. Paul (Stansifer) 16:04, 5 October 2011 (UTC)

I'm not sure I find this a compelling line of analysis, though I am myself suspicious of Kurzweil's singularity for another reason (viz.: most exponential processes, in the real world, hit practical problems pretty quickly, and go from being a hockey-stick to being an S-curve; it's not clear to me what the "resource" would be that exponential AI growth would run out of, but there is likely something out that that would serve as a cap, in the same way that quantum mechanics threatens to eventually put a stop to a strict definition of Moore's law). I find our mere decades of research to have been pretty fruitful so far (a decade is a long time to an individual human.. but not even that long; I can still remember what I was doing a decade ago quite vividly!). And the major difference between raising an AI researching human and writing an AI program is that once you have the program, you can duplicate it with a negligible amount of additional resources. The same cannot be said for the human! --Mr.98 (talk) 16:16, 5 October 2011 (UTC)

I think that an analogy can be made between artificial intelligence and the P vs. NP problem. We aren't very far along towards solving the problem (at least, Scott Aaronson says so, and I'll trust him), but we're making great progress on building better SAT solvers. A SAT solver can do a great job at model checking problems, but it shouldn't be taken as evidence that we're getting close to proving something about P and NP. Watson can do a great job at answering Jeopardy! questions, but it shouldn't be taken as evidence that we know anything about intelligence. This isn't to say that complexity theory and artificial intelligence/cognitive science researchers aren't accomplishing anything, just that their core problems are very large. Paul (Stansifer) 18:12, 5 October 2011 (UTC)

But P vs. NP isn't necessarily solvable in a rigorous way; in any case, it's an entirely artificial sort of problem (even if applications of it exist). Intelligence — at least what we call it — is not only solvable, but can emerge via natural processes. Nobody serious thinks SAT solvers or Watson are anything close to real artificial intelligence — except in the sense that raw memory and computational power does matter. At the moment, we're still a few orders of magnitude off from having the computing power to simulate, even in crude terms, a human brain. But our capacity for memory storage and our capacity for computing power still grows exponentially. In any case, there can't be anything too magical about intelligence if evolution can produce it. Evolution is clever, but it's not magical. --Mr.98 (talk) 00:10, 6 October 2011 (UTC)

The Science Fiction author Vernor Vinge has written several works dealing with the technological singularity, but most pertinently to the OP's query, his novel A Fire Upon the Deep explicitly portrays problems caused to advanced AIs by their exponentially self-increasing intelligence, and may therefore be of interest. {The poster formerly known as 87.81.230.195} 90.193.78.36 (talk) 17:23, 5 October 2011 (UTC)

That particular novel is probably not super-pertinent to the "runaway AI" question because of the stuff about the Unthinking Depths vs. the outer rim entities; but I highly, highly recommend the novel anyway. As to the original poster, you'll find that reference #1 in our technological singularity article points to this key 1993 paper by Vinge, which discusses the movement of humans away from center stage in the world, once runaway AI occurs. Comet Tuttle (talk) 20:31, 5 October 2011 (UTC)

For anyone who is interested, the best way to model the complexity of the proposed singularity (which is unlikely to happen the way it is described in this thread) is to take a look at astrobiology, particularly what is known about the emergence of life from inorganic matter, such as abiogenesis. You'll find a lot of answers there. Viriditas (talk) 22:33, 5 October 2011 (UTC)

Another fictional view on this "runaway AI" concept is "The Metamorphosis of Prime Intellect" (parts of this story are very NSFW, by the way), wherein the titular AI, "Prime Intellect" gains a level of understanding of how the Universe works on a low level, based on extrapolating upon available experimental data of a fictional phenomenon known as the "Correlation Effect", which sounds a bit like Quantum tunnelling. With this new found knowledge it is able to redesign and manufacture new and improved hardware for "itself", creating the exponential growth the OP mentions. The author doesn't give this process a name, however. I highly recommend the story, by the way, which has a Asimov-esque exploration of the Three Laws of Robotics as the core of the story. --Rixxin (talk) 10:52, 6 October 2011 (UTC)

Are there any reputable, free RAM "scrubbers" out there?

They would operate like (and hopefully better than) MemTurbo except that I would like a free utility. (MemTurbo is paid.)

Since my computer seems to keep running slow no matter what I try to close, minimize, etc., I need to find a good scrubber of RAM so that after it does its work, it runs faster than before it started.

So could someone please point me in the right direction? Thanks in advance! --70.179.161.22 (talk) 13:24, 5 October 2011 (UTC)

To point you in the right direction, "scrubbing" RAM does not, and can not, improve your computer's performance. Such programs are proverbial snake oil. Ensure that you are not running any unwanted programs that consume processor and memory resources. If your computer is still running below your expectations, you probably have to upgrade to newer hardware. Nimur (talk) 13:31, 5 October 2011 (UTC)

Come back, QEMM, all is forgiven. --Tagishsimon (talk) 15:05, 5 October 2011 (UTC)

MemTurbo gets 5 stars from cnet in the editor review, not sure if that says anything.

Windows 2000 had an option to optimize performance for either “Applications” or “Background Services". Don't know what a "scrubber" does, but predictive caching, pre-fetching, swap priorities based on user settings can all help. If they could do a better job than the OS system, I don't know. A fragmented swap file (if that still exists), or variable size can slow a pc down. Color settings, desktop themes, appearance, transparancy of windows, application settings (like limits for number and size of undo's and redo's, fonts to load, ..) there's so much that has an influence. DS Belgium (talk) 00:07, 6 October 2011 (UTC)

IOLO drive scrubber. — Preceding unsigned comment added by 74.178.177.94 (talk) 19:41, 5 October 2011 (UTC)

Don't think he wants to wipe his harddrive. Was this perhaps meant for the next section? DS Belgium (talk) 23:29, 5 October 2011 (UTC)

Dod Compliant wiping

Hello, I need to wipe a server and it needs to be DoD Compliant. After looking on my own I got lost and confused. Can some help me on what I need and places to get it. I got this old dell server and it needs to be DoD complaint wiped before I can use it for my own means. I have never donw anything with servers and was wondering if anyone could help

Thank you 152.27.56.61 (talk) 23:04, 5 October 2011 (UTC)

Have you read National Industrial Security Program#Data sanitization and checked out references 5 & 6? --Tagishsimon (talk) 23:08, 5 October 2011 (UTC)

See DBAN, but unless you've signed some legally binding agreement don't waste your time: one wipe with zeroes will suffice. ¦ Reisio (talk) 23:21, 5 October 2011 (UTC)

"DoD Compliant" is not specific enough; somebody needs to give you a specific requirement that you need to comply with. Different specifications exist for different types of data. If you aren't sure what you need to comply with, you should absolutely escalate the issue to your supervisors.

If somebody else gave you a server, and it actually did need to be wiped, it really was their responsibility to do so. Nimur (talk) 00:14, 6 October 2011 (UTC)

Yea, what a weird thing to do. "Here's a computer, but be sure to erase the files, because if you got your hands on them there would be trouble!" APL (talk) 03:16, 6 October 2011 (UTC)

I have worked for many different agencies under the DoD. All of them have specific wiping procedures. In one, there was a CD that you put in the computer and rebooted. The CD wiped the drive. All was good. In another, the drives had to be sent to a wiping center. When you get a used computer, you had to order a drive from the wiping center. In another, drives had to be shredded. If the user is looking for "DoD compliant", whatever agency he or she is with will have specific procedures for doing this. -- kainaw™ 12:51, 6 October 2011 (UTC)

Iolo drive scrubber. — Preceding unsigned comment added by 184.44.146.86 (talk) 16:31, 6 October 2011 (UTC)