|
COMPUTING AT CHAOS MANOR Jerry Pournelle Thursday, May 22, 2003 |
One thing was obvious at WinHEC: people aren't upgrading their machines because they don't think they need to, and the manufacturers are going nuts.
We have fast new machines. The only things they do that the last generation doesn't do is some modern games. There are niche tasks for fast machines, particularly in graphics processing, but for the most part are not things Good Enough?
So the inquiry is: What do we wish our machines would do that they do not do now? What do we want out of our machines, particularly out of out desktops? Obviously that's on a time scale. As Bob Thompson and I both said a year or so ago, we're in an odd situation: the machines we have are fast enough to do what we want, but they are way too slow to do some of the Artificial Intelligence things from the science fiction stories and films. So: given Moore's Law, and the machines get twice as fast every 18 months or so, and we are now up to 3 GHz and 800 MHz FSB and dual channel memory -- Question one: Given what we have available now or will have shortly, what do we wish the machines would do that they don't do? Question Two: We'll have an order of magnitude improvement in under a decade. What would we like THOSE machines to do? What software should the programmers be writing now for the next decade? At some point I will try to organize this. For the moment, here's a grab bag. We'll collect ideas until some organization becomes needed, or some sub topic springs out at us. These are in order of First in, First up; for latest see the bottom one.
|
|
I've some requests. 1) I'd really like to have better search functions. A super google on my own machine. I know I put it somewhere on the machine or machines...where is it? Help me find things on the web. Help me find things about my patients. Who haven't I treated with beta blockers, what am I forgetting to do. 2) Educational software for my children and for me. It is hard to learn things. The learning games available are overly simplistic. I'm struggling to "get" molecular cardiology...I'd pay to have educational software that helped me. Even having didactic lectures with on-line problem solving would help. I'd like to know more about architecture...fly me through examples of great architecture...let me drill down to get more information. Make the game "rich". Would a game for children that taught addition, subtraction, multiplication, and division in a role playing game be impossible to design? Mark Huth mhuthATcoldswim.com Be careful about reading health books. You may die of a misprint. twain == Hi. This is a response to a letter posted (recently?) asking why we don't have good educational software. The answer is that we're working on it, but it's a
hard slog. There actually *is* good stuff for teaching elementary
mathematics, for example AnimalWatch -- http://sophia.cs.umass.edu/AWE/
-- and really good intelligent-tutor based curricula for middle school
and high school level math (note: I'm a graduate student with the people
who did this stuff) -- see http://www.cogsci.northwestern.edu/ There's not much that's good for advanced subjects like molecular cardiology, though. The reason it's so slow is that the technology isn't the hard part of designing these systems. The hard part is studying and modeling student cognition and tutor behavior to the level of detail necessary to create sophisticated educational systems. I think this problem is general to the next generation of user interfaces. Sure, there are things we'll be able to do that we can't do now when we have super-powerful nanoscale computers; but our ability to understand and design things to the needs of human cognition is the real bottleneck, not hardware speed and power. Ryan Shaun Baker PhD Candidate Human-Computer Interaction Institute Carnegie Mellon University ========================= Information appliances "good enough" to index and cross link all kinds of info we squirrel away. Zoot shows the way now, Chandler may do it better. So then automatic backups to media so we don't lose what we once had. Some kind of socio-historical simulation like Hesse's Glass Bead Game. A way to see simultaneous globe wide changes and calculate what-if scenarios. Means to not forget important factors, and to overcome the "all other things being equal" evasion. Joe O'Laughlin Jerry, (1) In response to Mr. Huth: I developed conceptual designs for such games over a decade ago. Proposed them to my then department manager (remember, this was Teledyne Brown, back when it was still reeling from Gen Graham's comments about "taking over" SDI). His response was something along the lines of, "So, you want to do computer gaming now." Period. No follow-up, funding, or interest in commercial software there, and so I let it slide. The ideas are languishing, and I've never seen anything close. Any takers that we might mate my current company with? (2) This topic plays to a thought I've had for some time; specifically and to wit: the computer revolution has NOT yet arrived. So far, the (digital) computer has been an EVOLUTIONARY tool; we are doing the same things we did before we had computers, some differently, and with greater data management flexibilty, and greater speed, and in some cases perhaps not as well. But I can't think of a single example of anything that we are doing today with (digital) computers that we didn't do fifty years ago without (digital) computers. Note that I AM hedging by the use of the term digital, which is what most people mean by computer, because we were doing most of those functions with analog computers, both electronic and mechanical, and other analog technologies such as pressed LP records traceable to Edison. (3) The changes I see in the short term that would benefit from Moore's law: truly real-time voice recognition software with either typing or real-time translation capability, at least to known languages. (The "universal translator" beloved of Trekkers is at least one, if not two, generations away.) Fully functional net-capable videophones with 20 - 32 screens/second of VGA-or better quality graphics is probably less a function of CPU speed than of connectivity, although for videophone or videoconferencing applications 95:1 data compression might be possible to accelerate that day. Readily accessible household integration -- send an E-mail to start dinner cooking -- with corresponding household robotics and security functions (shades of "Pandora's Box," "Stranger," and most particularly "The Door into Summer"). Biometrics. Medical history data mining that turns personal records into real-time epidemological data bases at the level of individual diet and resulting blood chemistry changes across whole populations -- without compromising patient privacy or patient-doctor confidentiality -- and leads to AMA-certified and accepted diagnostic assistance software and a new generation of medical analytical technologies, indubitably the first step towards Mr. Niven's Autodocs. Automated (neural net?) tools to codify all of physics back to Page 1 of Issue 1 of Volume 1 of the Physical Review - or better yet, the Monthly Notices of the Royal Society, and ditto for chemistry and biology, with the ability to point up holes and allow real-time electronic access to the whole history of study of a subject at the desktop. The neural interface that turns this into the Norlamminian "mechanical educator" of Dr. Smith's classic "Skylark Three" -- or the electromechanical teleaphy of Mr. Robinson's equally classic "Mindkiller," with all of the protectors of privacy and of persons --and of sanity -- that Mr. Robinson noted must be necessary. I think this is enough suggestions to get you started. Jimwoosley@aol.com ===================== I want nothing short of a star trek holodeck out of my computer. I doubt I'll ever see the matter-interaction part of that in my lifetime, but the rest is mainly clever software using "fast enough" hardware. I also want my current computer's capabilities to be integrated into a 20" (approx size) transparent flexible (rollable and/or foldable) sheet that can be stuck to or wrapped around any surface. That way I can turn any surface into a fully functioning computer. I want a wearable computer device integrating a GPS, wearable semi-transparent display (integrated into glasses or contacts), rangefinder/surface scanner, multi-spectral camera, inertial nav system, and an object database. The functionality must include the ability to store and recall "tag" information on any object within sight, overlay that info onto the real world as seen through the display, give alerts on hazards (objects moving rapidly in my direction), overlay the multi-spectral image through the display so I can get heat/UV/etc images overlayed on top of objects, object/face recognition, and of course get directions to any location or address on the planet. The database should allow for filterable info such as items of historical interest, business names and contact information, names of people, even the going price of a gallon of milk when I'm in the supermarket. It needs a virtual keyboard or subvocal speech recognition so I can add tags to objects within my field of view. It should run for a week without needing more power, and the power source should be be available as easily as one gets a drink of water. My wearable device should also be "connected" to any future internet, have the ability to securely link up or deny links to and from any other similiar device, and should have integrated directional speakers so only I can hear it's audio output without having to wear bulky headphones or even earbuds. All I want for my desktop is the ability to freely manipulate any series of bits stored on any type of media owned or possessed by me. Microsoft, the RIAA, and the MPAA are the only ones trying to take that basic capability away. My college 386 back in 1991 is more suitable for bit twiddling than the future platforms envisioned by microsoft, so my current desktop and win98SE installation sits right next to my rifle on my cold-dead-fingers list. (WinXP sits next to turbotax on my 10-ft-pole list.) Sean Long I said it when Microsoft came out with the new keyboard with 105 keys. They added (among others) a Windows key. What they needed (and what we still need) was a "Do what I mean" key. Aside from that, I want a search engine for the internet that lets me do a search, and if I get too many hits, lets me do a further search on just the results of the previous search. How about rental software. I need a program for one job or project but it is not worth spending $100, $1000, $10,000 for it. I log on to a service and pay by the hour to use it. How about a library of old software. There is a specific program that does just the job I need but I don't have it any more or it will not run properly on my current machine. Log on and pay to use it for one job. I'll think of other stuff eventually. Patrick A. Hoage Emphasis added by editor. I would be astonished if there were not software out there that does that (search only among previous hits) but I don't think of any off hand. And doesn't the .NET concept attempt the rental software idea? It certainly could. And several readers point out: If you scroll to the bottom of a page of Google results, there is an option "search within results". The moral of this story is that often our machines can in fact do the things we want, but they do a lousy job of making us aware of that. It happens to me daily.
==== Jerry, What I want my computers (ALL of them) to do, is to simply GROW UP! I am sick and tired of prompt after prompt after prompt! For example; Appleworks - simple find and/or change text.. (and NONE of the four possible commands can be completed with return/enter - you MUST use the mouse!) Change All returns: "The Change feature is not undoable." with OK or Cancel OK returns: "x occurance(s) replaced." with OK (ONLY!) Photoshop 7.x can't remember where you last saved an edited photo, so EVERY save requires multiple clicks, prompts, and manual commands that are ALL unneeded. (hey Adobe, go try Graphic Converter and learn HOW to handle saving!) My Roland VS Recorder not only asks if you are sure you want your command carried out (huh?), it then asks "Are you really sure?" And if you are burning a CDR it takes ANOTHER FOUR prompts! SUMMARY: we need a Constitutional Amendment or something nearly that strong, MANDATING that ALL software gives the user the CHOICE of turning off ALL prompts! But of course what we would get when initiating that choice would probably go like this.. Are you sure? Ok or Cancel Are you really sure? Ok or Cancel You idiot, you can't function without prompts! True or Agree or Cancel or I'll Risk It Ha Ha Ha! You didnt' actually think we'd let you turn us off did you? No or Damn It Turning off all prompts will void your purchase and require buying this software again. (Gotch!) Lee -------------- Calculus concretion in rotary transition gleans slight bryophytic accretion. We can all agree... Subject: New Tasks For More Powerful Computers Here in Florida are a number of non-native invasive plant and animal species causing big problems with the original ecosystems. I assume at some point genome mapping will become inexpensive. I would like to see this technology applied to mapping out these non-native species and then using our future computers via simulations to exploit distinguishing characteristics so as to design highly selective pesticides and chemicals that only take out these non-native species. I assume a great deal of computing power would be helpful towards this end. Mike Cheek Tallahassee, Florida This one opens a whole field of speculations. It will, I think, be a while before we have small automated desktop gene splicers: but not forever. In response to 'What Do We Want These Macines To Do?' Dr. Pournelle, To be honest, while computers exist that make our life easier, I think the true purpose of the computer is entertainment. If not, why are there so many more games out there than any other kinds of software. People even pay monthly fees to play some games, Everquest being my favorite. In truth, it is the entertainment market that drives the hardware industry. Look at the current battle between nVidia and ATI. In quick succession nVidia releases the GeforceFX cards and updates them. In rebuttal ATI releases the Radeon 9800 Pro, and then, in response to the Geforce FX cards, updates it with a 256MB DDR-2 card. Why? It's all about FPS, Frames Per Second. Whether you are killing Nazis, mutants, or aliens you want it to be the cleanest, smoothest animation possible. Are there other uses for these products, of course. Would these products have been developed without the gaming industry, yes, but at a much more sedate pace. People upgrade their systems, not because they need to, but because they need a little performance boost. Remember, as processors get faster applications get larger, and are expected to do more. Finally, what do I want these machines to do? I want to sit down and fire up my copy of Mechwarrior 8 and really feel I'm in a battlemech. I want to fly an F-22 in a dogfight and see the cockpit around me, to reach out and stroke the controls and have them respond. In an online RPG, Role-Playing Game, I want to interact with an NPC, Non-Player Character, and not be able to tell if it is an AI or not. I want a machine to pass the Turing test. I want to see an android so humaniform that people get scared. As far as the gaming aspects of what I said, I realize that the major issues come from the HIDs, Human Interface Devices. To do what I want, we need to trick the brain into thinking that the body is actually in motion. This leads in two totally different directions. The first direction is virtual reality suits. Suits that are capable of inducing sensations in the nerves of the body. The other direction has the possibilty of turning into a very controversial issue. I am talking about direct stimulation of the brain through implants. I know that you have touched on the Cyberpunk genre with the Saurons, but they are definitely not punk. They are however frought with cybernetics and nanotech. Why controversial? Implants. The religious right will start saying things like, "Mark of the Beast." People not necessarily associated with the religious right will wonder if we are meddling with things we shouldn't. There are going to be lawsuits from operations gone awry. Now, to be honest, I fall on the religious right myself. I even hold to the tenets of Christianity. I do not know how I'm going to side on this argument. There are some other possibilities for this technology, nerve inductions fields and the like, but I think these are the most likely in the short run. These two technologies at least have a basis in reality. Look at the man who regained his sight, at least partially, through implants in his optic nerve. We are at a crux and the issues are only going get more grey. Well, such are my thoughts, Doug
Indeed.
This is my second letter in response to What Do We Want These Machines To Do. I want them to live. I want Mr. Data in every household. If there is a dirty, nasty, repetitive task I want a machine to do it, and appear cheerful while doing so. Man has already exhibited the ability to become attached to their computers. Imagine if it becomes that much easier to anthropomorphize them. Every family a butler or maid. Mechanized babysitters. The uses are endless. There is another reason, though. I want to watch the social upheaval. I can already see the android activist movement. Hundred's of unwashed bodies marching the streets of America with signs and placards, "ANDROIDS ARE HUMAN TOO," "FREE THE 'DROIDS." Then there would be the TV spots. "This is my friend Bob. He's an android." Don't believe it will happen? These are Americans we are talking about. We are some of the best at jumping on bandwagons. If most of America keeps a level head, or can be made to care it will go nowhere. Can we count on a large portion of the American population to care? Well, to look at recent history, and by recent I mean the last forty years, probably not. I can already see the battered android shelters and the Android protection laws. I also see the entrance of androids into the vice arena. Imagine, bordellos staffed completely with androids programmed to meet every desire possible. Let's not forget child-like androids for that particular set, and a whole new set of laws to go along. This will also bring the church into the issue, not that they wouldn't already be there. The more radical of the Christian sects, and I can see Muslim there also, will be decrying humaniform androids as Godless and evil. Once we start seeing mechanical sex toys the controversy will be over whether sex with an android is adultery. "Well, she's not a real woman." This argument will be made by husbands and argued over by clergy. This brings us to the Terminator scenario. Do I think this will or can happen? No, I do not. While the ability to produce such machines is going to come about, man's ability to react after being burned will come into play. One rogue AI, and there will be controls placed on machine intelligences that will take care of the issue. It will take something of the sort to wake humanity up though. We have a history of looking the other way until our backside gets bitten. Even then, the android rights movement will get into the picture. I can see the ACLU spokesman on television now, "Should all androids be judged by the actions of one?" Basically, I want androids to become as anthropomorphic as possible. I want Artificial Intelligences to be able to, or at least appear as able to think and make decisions, or more precisely make judgment calls, to make a decision where the arguments made are ambiguous. I want this for two reasons. Number one, the utility. Androids could be a liberating force. Second, I want to sit back and watch the next act in the American circus. Ave atque vale, Douglas Knapp That is certainly the ultimate wish. You remind me that this is your second letter, which reminds me to say that there is no great guiding principle here: I get lots of mail, far more than I can comment on or even post, and when it comes has much to do with what happens: sometimes a lot of good stuff comes when I have many other things to do, and then the mail is put where I can get to it when I have time, and I generally don't have time, and... In any event I went back and found your original and that is up here too. Anyway I think it safe to say we won't have real AI for a while. What can we do with the machines we have now, or by Moore's Law will have shortly? But I agree, Isaac's Robots, or those who served the Websters, would be very useful...
=
|