josh@klaatu.rutgers.edu (J Storrs Hall) (02/22/90)
Well, not too long ago, I sent the following to comp.arch: what 1980 1990 extrapolation 2000 processor pins 64 256 x4 1k clock 8MHz 32MHz x4 128MHz "Devices" 50k 1M x20 20M mips 1 12 x12 144 memory(dram) bits/chip 16k 1M x64 64M speed 240ns 80ns /4 20ns and got a flood of helpful responses, agreeing or disagreeing about various points. I've been updating it since, on the basis of that and other information. Here's the current best guess: (and allow me to point out that I changed the 1990 mips to 20 *before* the 68040 announcement appeared...) what 1980 1990 extrapolation 2000 processor pins 64 256 x4 1k clock 8MHz 32MHz x4 128MHz "Devices" 50k 1M x20 20M mips 1 20 x20 400 memory(dram) bits/chip 16k 4M x256 1G speed 240ns 80ns /3 25ns An interesting thing is that one fellow objected to the figures as too tame, because they were a "linear" extrapolation! Trendline forecasting like this is of VERY limited usefulness, and that is just as true of the numerically sophisticated forms like the "Club of Rome" models. Basically you're saying, things are going to stay the same as they are now, only more so. I would claim that the best method of forecasting is still to take someone who is a sharp social and technological observer and have him apply his whole mind to the problem, using any and all analytical tools available. One of my favorite scenes from science fiction is in "Door into Summer", by Heinlein (a SS&TO, see above). The protagonist is talking to a workman who is crushing cars in a big machine. It eventuates that they don't have engines. They are being built, and crushed, simply to provide jobs for the builders and crushers [quoted from old and unreliable memory]: "But why no engines?" "Well, it doesn't make sense to put in an engine if it's just going to be crushed, does it?" "But it doesn't make sense to build the car in the first place if it's only going to be crushed!" "Buddy, you just don't understand economics at all, do you?" --JoSH
macdon@ists.ists.ca (Rick MacDonald) (02/23/90)
In <Feb.21.17.13.14.1990.1539@klaatu.rutgers.edu>, J. Storrs Hall gives an interesting example of predicting processor capability in the year 2000, and concludes: >Trendline forecasting like this is of VERY limited usefulness, >and that is just as true of the numerically sophisticated forms >like the "Club of Rome" models. Basically you're saying, things >are going to stay the same as they are now, only more so. >I would claim that the best method of forecasting is still >to take someone who is a sharp social and technological observer >and have him apply his whole mind to the problem, using any >and all analytical tools available. If I were in the business of designing microprocessors, and doing technology forecasting to decide what R&D to do, I would be very concerned about keeping track of advances in optical and molecular computing and the like. Or perhaps, breakthroughs in neural nets will lead to a whole new processor technology, resulting in a limited need for microprocessors as we know them by the year 2000. These would be examples of 'blindsiding'. My point is that you have to be aware of all the technologies that could affect your business, and think about their future impact. That is a tough job. One thing that's clear to me is that a single individual is going to have a much harder time doing forecasting like this than the right group of individuals. What kind of individuals should that group consist of and how should they do their job? More importantly, how do they convince anybody that their forecasts are worth acting on? Rick MacDonald Spar Aerospace Ltd. macdon@ists.ists.ca -- Rick MacDonald Spar Aerospace Ltd. macdon@ists.ists.ca
bill@videovax.tv.tek.com (William K. McFadden) (02/24/90)
In article <Feb.21.17.13.14.1990.1539@klaatu.rutgers.edu> josh@klaatu.rutgers.edu (J Storrs Hall) writes: > >what 1980 1990 extrapolation 2000 >processor >memory(dram) >bits/chip 16k 4M x256 1G This is the only one of your figures I've been tracking over the years. My personal rule-of-thumb says memory capacity quadruples about every three years. I'm not sure of the first few dates, but I predict the following trend: 1978 16K 1981 64K 1984 256K 1987 1M 1990 4M 1993 16M 1996 64M 1999 256M 2002 1G 2005 4G 2007 16G 2010 64G These dates are those for mass production, not first silicon in the lab. Also, they are approximate and of course speculative, especially the latter ones. However, last year I saw an overhead from a major DRAM manufacturer predicting size increases (including the 64G size!) that almost exactly matched my rule of thumb. Remember, this advice is worth every cent you paid for it! -- Bill McFadden Tektronix, Inc. P.O. Box 500 MS 58-639 Beaverton, OR 97077 bill@videovax.tv.tek.com, {hplabs,uw-beaver,decvax}!tektronix!videovax!bill Phone: (503) 627-6920 "The biggest difference between developing a missle component and a toy is the 'cost constraint.'" -- John Anderson, Engineer, TI
weber@popvax.harvard.edu (R09400@RPWeber) (02/24/90)
The main problem of forecasting is uncertainty. Trends are unhelpful
in dealing with uncertainty since they lead to roadsigns or milestones
that tell you if something has or has not happened. For example,
we all know that the cost of computing has been declining and is likely
to do so. What does knowing the trend help us to forecast or understand?
I would suggest nothing. However, if one were to think about the
future in terms of events, then an event such as, Apple will introduce
a "throwaway mac II" for $ 600 in 1994 , is an event that reflects
the declining cost of computing. We can then ask, did this event
happen or not? The future, then, is made of sequences of events.
These ideas are elaborated in an article that appeared last November
27 in Computerworld, authored by Jim Herman, Patricia Seybold, and me.
Anyone intested in forecasting or the future of computing and communications
technology should look at this article.
Regards
Bob Weber
-----------------------------------------------------------------------
Robert Philip Weber, Ph.D. | Phone: (617) 495-3744
Senior Consultant | Fax: (617) 495-0750
Academic and Planning Services |
Division |
Office For Information Technology| Internet: weber@popvax.harvard.edu
Harvard University | Bitnet: Weber@Harvarda
50 Church Street |
Cambridge MA 02138 |weber@popvax.harvard.edu (R09400@RPWeber) (02/24/90)
Oops (not a good unix hacker, yet), it should have read,
trends do NOT lead to signposts or milestones.
Bob Weber
-----------------------------------------------------------------------
Robert Philip Weber, Ph.D. | Phone: (617) 495-3744
Senior Consultant | Fax: (617) 495-0750
Academic and Planning Services |
Division |
Office For Information Technology| Internet: weber@popvax.harvard.edu
Harvard University | Bitnet: Weber@Harvarda
50 Church Street |
Cambridge MA 02138 |bzs@world.std.com (Barry Shein) (02/24/90)
From: uunet!husc6.harvard.edu!popvax!weber (R09400@RPWeber) >The main problem of forecasting is uncertainty. Trends are unhelpful >in dealing with uncertainty since they lead to roadsigns or milestones >that tell you if something has or has not happened. For example, >we all know that the cost of computing has been declining and is likely >to do so. What does knowing the trend help us to forecast or understand? >I would suggest nothing. However, if one were to think about the >future in terms of events, then an event such as, Apple will introduce >a "throwaway mac II" for $ 600 in 1994 , is an event that reflects >the declining cost of computing. We can then ask, did this event >happen or not? The future, then, is made of sequences of events. That's very well said, and subtle enough that I hope most people read it again. The first question which comes to my mind is, is this an either/or situation? Don't the trend lines at least provide an envelope of possibility to place events in, a probability space basically. Something lying outside of the extrapolation boundary implies a "catastrophic precursor", such as a new chip technology allowing a sudden 100x speed-up, so we go back and look for that. -Barry Shein Software Tool & Die | {xylogics,uunet}!world!bzs | bzs@world.std.com Purveyors to the Trade | Voice: 617-739-0202 | Login: 617-739-WRLD
weber@popvax.harvard.edu (R09400@RPWeber) (02/24/90)
In article <9002240412.AA14891@world.std.com> bzs@world.std.com (Barry Shein) writes: >The first question which comes to my mind is, is this an either/or >situation? Don't the trend lines at least provide an envelope of >possibility to place events in, a probability space basically. > >Something lying outside of the extrapolation boundary implies a >"catastrophic precursor", such as a new chip technology allowing a >sudden 100x speed-up, so we go back and look for that. > The problem with probability space is that if you grant the premise that the future occurs because of some event sequence (rather than because of some sent of trends (even thought the former reflect the more likely than any other future. For example, if you have 40 events in a sequence leading to some outcome, and each event is highly probable, say .9 likely, then the probability of this particular future is .9 to the 40th power, or .017. This tells us that the world is highly contingent. One goal should be to give up the notion of one particular likely outcome and think in terms of alternative event sequences and alternative outcomes. Regards Bob ------------------------------------------------------------------ Bob Weber | Voice (617) 495-3744 Sr. Technology Consultant | Fax (617) 495-0750 Harvard University/OIT | Bitnet: weber@harvarda 50 Church St., 4th Floor | Internet: weber@popvax.harvard.edu Cambridge, MA 02138 ------------------------------------------------------------------
v291nhtp@ubvms.cc.buffalo.edu (02/24/90)
In article <5462@ists.ists.ca>, macdon@ists.ists.ca (Rick MacDonald) writes: > > In <Feb.21.17.13.14.1990.1539@klaatu.rutgers.edu>, J. Storrs Hall gives an > interesting example of predicting processor capability in the year 2000, > and concludes: >>I would claim that the best method of forecasting is still >>to take someone who is a sharp social and technological observer >>and have him apply his whole mind to the problem, using any >>and all analytical tools available. This is exactly what R. Buckminster Fuller did throughout his life. (He is know [among other things] for inventing the geodesic dome.) His [and my] philosophy entailed NOT speciallizing in any one area, but standing back and getting "The Big Picture." Taking a holistic, omni-interrelated look at the universe, and seeing how things fit together. > My point is that you have to be aware of all the technologies that could > affect your business, and think about their future impact. That is a tough > job. True. But SOMEBODY'S gotta do it! :) (Actually a LOT of somebodys should do it, but that's another story.....) > One thing that's clear to me is that a single individual is going to have a > much harder time doing forecasting like this than the right group of > individuals. True. Wanna help us out? I'm already part of an informal group doing that. (See below.) > What kind of individuals should that group consist of and how should they > do their job? All kinds of individual from all walks of life. Everyone has SOMETHING to offer. The whole key is to have the desire and imagination to dare to dream! That's what seems to bond us together. > More importantly, how do they convince anybody that their > forecasts are worth acting on? Ah! Now we're getting to the crux of the matter! This is one of the biggest walls I keep slamming into. It's tricky, and I can't quite figure it out. I think it has something to do with having more money than most small countries, so you can do anything you want, and not be liable to a bank for it. ;^) If the idea is good enough, and is "right," then people will spontaneously adopt it and discard the older, less efficient way. The key is bringing said idea to fruition. Like I said, I'm already a part of this. I have a Listserv on BITNET called GEODESIC, which is a sort of free-form discussion on various technologies, from space shuttle fuel-cell power systems, to toilets ( ;) ), and how they impact and interrelate to improve the standard of living for all humans on the planet, through intelligent design and proper application of the principles of the Universe. I'm in the process of peering the list to the BRAND NEW NEWSGROUP (what a plug.) ***BIT.LISTSERV.GEODESIC*** (Outrageous! How can people get away with such tawdry advertising schemes?!?) ;^) Check it out! I think you might like it! (IF we can get it running! :) ) > Rick MacDonald > Spar Aerospace Ltd. > macdon@ists.ists.ca -Patrick Salsbury U. of Buffalo, NY V291NHTP@UBVMS.BITNET
KPURCELL@liverpool.ac.UK (Kevin 'fractal' Purcell) (02/27/90)
On 23 Feb 90 20:29:46 GMT William K. McFadden (zephyr.ens.tek.com!tekfdi!videovax!bill@edu.washington.cs.beaver) said: >In article <Feb.21.17.13.14.1990.1539@klaatu.rutgers.edu> > josh@klaatu.rutgers.edu (J Storrs Hall) writes: >> >>what 1980 1990 extrapolation 2000 >>processor >>memory(dram) >>bits/chip 16k 4M x256 1G > >This is the only one of your figures I've been tracking over the years. My >personal rule-of-thumb says memory capacity quadruples about every three years. >I'm not sure of the first few dates, but I predict the following trend: > >1978 16K >1981 64K >1984 256K >1987 1M >1990 4M >1993 16M >1996 64M At this point there will be a breakpoint as standard ULSI production techniques will have to be replaced with something better, e.g using synchrotron radiation to move to smaller wavelenghths for making the patterns in the resist coupled with better matching of masks to wafers. Going beyond 64Mbytes features will need to be less than 100nm then we real start to hit FUNDAMENTAL size limits such as the number of electrons you can keep in a memory cell decreases and (of course) quantum effects. Around the millenium the paradigm of computing will have changed from this pattern. I would expect to see memory more closely coupled to the processing elements, with von neumann processors drifting out of fashion. We may even have got somewhere with nanotechnology! >1999 256M >2002 1G >2005 4G >2007 16G >2010 64G > >Bill McFadden Tektronix, Inc. P.O. Box 500 MS 58-639 Beaverton, OR 97077 =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= Kevin 'fractal' Purcell ...................... kpurcell @ liverpool.ac.uk Surface Science Centre, Liverpool University =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= "It is time that knowledge became more accessible to those to whom it properly belongs" -- James Burke, _The Day the Universe Changed_
matheson@portia.Stanford.EDU (David Matheson) (02/27/90)
In article <1936@husc6.harvard.edu> weber@popvax.harvard.edu (Bob Weber) writes: >The problem with probability space is that if you grant the premise >that the future occurs because of some event sequence (rather than >because of some sent of trends (even thought the former reflect the >more likely than any other future. For example, if you have 40 events >in a sequence leading to some outcome, and each event is highly probable, >say .9 likely, then the probability of this particular future is >.9 to the 40th power, or .017. This tells us that the world is highly >contingent. One goal should be to give up the notion of one particular >likely outcome and think in terms of alternative event sequences and >alternative outcomes. > >Regards >Bob Although I wholly agree with Bob's event notions for thinking about the technology of the future, I would have to take issue with the notion that probability is not very helpful. When making decisions (i.e. committing $ now) in anticipation of the future state of technology, it is critically important to know which scenarios (or sequences of events) are more plausible than others. What I have typically done is to: 1. Construct event definitions to serve as the progress of technology 2. Select representative sequences to serve as possible futures. 3. Assess a probability distribution over future states, using the representatives as proxies. Using models to help people represent their information, current decision tree technology can reach around 2,000 such scenarios. 4. Although the probability of any particular scenario is small (as Bob points out), one does not need to know the future in order to make a decision today. Rather, the entire collection of future possibilities should be taken into account. I think my work has a somewhat different focus that typical technology forecasting. Some want to forecast technology for the sake of forecasting technology (a slight charactature). I always have a specific decision in mind. Thus my event definintions, modelling, and so are all constructed so as to make the correct decisino as clear as possible. David Matheson Stanford University -- ______________________________________________________________ David Matheson matheson@portia.stanford.edu 376 College #5, Palo Alto, CA 94306-1545 (415) 328-3515
gary@racine.ACA.MCC.COM (Gary Knight) (03/01/90)
For those interested in retrospective analysis of technology forecasting efforts, pick up a copy of Steve Schnaars book, MEGAMISTAKES. You'll find it enlightening. I spent an afternoon talking with Schnaars after the book came out, and came away with a much better understanding of why TFing has utterly failed when held to any scientific standard. Schnaars' advice was (1) watch your assumptions, and (2) use common sense, and forget all the bells and whistles. Rather than trying to forecast (which has lots of things going against it, including (1) high error rate, (2) assuming that tomorrow is going to be just like yesterday, (3) people actually believe forecasts and get complacent as a result, (4) lack of a global, qualitative approach, it's probably preferable to work from scenarios -- i.e., determine the future state you want to achieve, block out a path to reach it, identify all the events that might occur and present obstacles or opportunities, and then plan out contingencies for such events. This way you control the future, in a sense, rather than futilely trying to predict it. For what it's worth . . . . And "Hi" Paul -- when are we going to start talking about this stuff again?!?! -- Gary Knight Technology Foresight Analyst MCC gary@mcc.com 3500 W. Balcones Center Dr. phone 512/338-3694 Austin TX 78759-6509 fax 512/338-3898
bzs@world.std.com (Barry Shein) (03/02/90)
Good idea, let's get books listed...
AUTHOR Smith, Douglas K.
TITLE Fumbling the future : how Xerox invented, then ignored, the
first personal computer / Douglas K. Smith and Robert C.
Alexander.
PUBLICATION 1st ed. New York : W. Morrow, c1988.
I couldn't get an ISBN out of the catalog, possibly my fault, but the
book is current and should be at your local popular title bookstore.
I've read it, it's a fun story although I don't know what others
involved in Xerox during those years have said about the book's
accuracy. Definitely non-technical and geared towards business types.
-Barry Shein
Software Tool & Die | {xylogics,uunet}!world!bzs | bzs@world.std.com
Purveyors to the Trade | Voice: 617-739-0202 | Login: 617-739-WRLDken@cs.rochester.edu (Ken Yap) (03/02/90)
|AUTHOR Smith, Douglas K. |TITLE Fumbling the future : how Xerox invented, then ignored, the | first personal computer / Douglas K. Smith and Robert C. | Alexander. |PUBLICATION 1st ed. New York : W. Morrow, c1988. | |I couldn't get an ISBN out of the catalog, possibly my fault, but the |book is current and should be at your local popular title bookstore. | |I've read it, it's a fun story although I don't know what others |involved in Xerox during those years have said about the book's |accuracy. Definitely non-technical and geared towards business types. A funny story. When I borrowed this book out of our library, I found a postit note inside saying "This story is TRUE. Signed: a Xerox person". (Xerox has several sites in Rochester). Take this story however you will. :-)
dana@atexnet.UUCP (Dana Burns) (03/15/90)
In article <34918@cci632.UUCP> jct@ccird3.UUCP (John Thompson) writes: >In article <84.25e65edf@ubvms.cc.buffalo.edu> v291nhtp@ubvms.cc.buffalo.edu writes: >>;^) If the idea is good enough, and is "right," then people will spontaneously >>adopt it and discard the older, less efficient way. The key is bringing said >>idea to fruition. > >How do you define "right"? My obsevation of most new ideas is that most >people will resist them, not on any "logical" ( a word I always find to >be suspect ) grounds, but basically for the following two reasons : > > 1) It's not the way our ancestors/fathers/mentors did it. > 2) I don't understand it, therefore it won't work "for me". > >Efficency has little to do with it. It seems to be that the older idea is >"time proven" therefore comfortable. All real advances in this world (IMO) >seem to have come about because one person or group has had a vision of >a "better" way and forced the world to listen. > I don't usually post, but here's a story you guys might enjoy: Someone I know went to work for this shoe-machine company. Now, shoes manufacturing is a very complex business due to all the sizes and shapes etc... It is also an industry steeped in tradition, and as an example, my old coworker related this tidbit: Seems there is this mechanical device that traces a prototype onto patterns from which tooling is made for the manufacturing equipment. It was invented in the 1800's and is still in use. This shoe-machine company has created a computer-system for automatically generating these patterns from a CAD system. All the patterns can be generated based on these exact tables and the system works great. Just got to sell it now. Well, it turns out that the old mechanical thing (wish I could remember what they call it...) has some inaccuracies in it due to mechanical joint limitations (or some such ...). To assess the new computer-based system (which lack the inaccuracies), they compare the patterns generated TO THE ONES GENERATED BY THE OLD MECHANICAL THINGY! When they don't match very well, they say: "this new fangled thing doesn't work for sh*t". When the sales people try to explain that the problem lies in the mechanical inaccuracies, they get: "son, this things been doing the job for A HUNDRED YEARS!" and they get thrown out the door. So, back to the drawing board. Now the company is spending mucho development studying the old thingy, so they can build the same inaccuracies into their computer. //// |Dana Burns <dana@bottle.atexnet.UUCP> ------HHHHHHHHHH |ATEX Publishing Systems \\\\ |"opinions mine" -- //// |Dana Burns <dana@bottle.atexnet.UUCP> ------HHHHHHHHHH |ATEX Publishing Systems \\\\ |"opinions mine"