ACM 97 Laws of Predictions Gordon Bell Microsoft The Folly 1.0.
Post on 26-Mar-2015
212 Views
Preview:
Transcript
ACM 97
Laws
of Predictions
Gordon Bell
Microsoft
The Folly 1.0
ACM 97
ACM 97ACM 97THE NEXT 50 YEARS OF COMPUTING
ACM 97
ACM 97ACM 97
Copyright Copyright 1997 ACM, Association for Computing 1997 ACM, Association for Computing
The files on this disk or server have been provided by ACM. The files distributed by this server have been The files on this disk or server have been provided by ACM. The files distributed by this server have been provided by ACM. Copyright and all rights therein are maintained by ACM. It is understood that all persons provided by ACM. Copyright and all rights therein are maintained by ACM. It is understood that all persons copying this information will adhere to the terms and constraints invoked by ACM’s copyright. These works copying this information will adhere to the terms and constraints invoked by ACM’s copyright. These works may not be reposted without the explicit permission of ACM. Reuse and/or reposting for noncommercial may not be reposted without the explicit permission of ACM. Reuse and/or reposting for noncommercial classroom use is permitted. Questions regarding usage rights and permissions may be addressed to: classroom use is permitted. Questions regarding usage rights and permissions may be addressed to: permissions@acm.orgpermissions@acm.org
THE NEXT 50 YEARS OF COMPUTINGTHE NEXT 50 YEARS OF COMPUTING
ACM 97James BurkeJames Burke
Master of CeremoniesMaster of Ceremonies
ACM 97
ACM 97
ACM 97
ACM 97GORDON BELL
ACM 97
Laws
of Predictions
Gordon Bell
Microsoft
The Folly 1.0
ACM 97
I think there is a world I think there is a world
market for maybe five market for maybe five
computers.computers.
““
”” Thomas Watson Senior, Chairman of IBM, 1943
ACM 97ACM 97
ACM 97ACM 97ACM 97
Harvard Mark I aka IBM ASCC
ACM 97
Predictions require Predictions require some history.some history.
The computer hadn’t been invented.
Watson’s prediction held for 10 years.
You need history to predict.ACM 97ACM 97
ACM 97
ACM 97
Navy Delphi Panel1969
Card readers will peak at 1500 Card readers will peak at 1500 cards per minute by 1974 and cards per minute by 1974 and then their use will decline.then their use will decline.
Advances in cores, wire, and thin Advances in cores, wire, and thin film will provide large memories film will provide large memories with one million words by 1976.with one million words by 1976.
““ ””
”” ““
ACM 97ACM 97
ACM 97
““ ””
I’ll bet we’ll be I’ll bet we’ll be
manufacturing cores manufacturing cores
in 1980.in 1980.
H. LamireVP Manufacturing, Digital
1975
ACM 97ACM 97
ACM 97
A technology can come A technology can come from nowhere and wipe from nowhere and wipe you out!you out!
MOS memories and micros were introduced in 1972.
By 1976, MOS memories wiped out core memories.
ACM 97ACM 97
ACM 97
Moore’s Law 60%/yr. Memory -- 4 x size every 3 years
10 G
1 G
100 M
10 M
1 M
100 K
10 K
1 K
0.1 K1970 1980 1990 2000 2010
ACM 97ACM 97
ACM 97
““
””
Moore’s Law predicts 2.55 Moore’s Law predicts 2.55 PetaBytes/chip in 30 years. PetaBytes/chip in 30 years.
On 1/97 a U.of Minn. team On 1/97 a U.of Minn. team stored one electron in a stored one electron in a cell at this density.cell at this density.
Nathan MyhrvoldMicrosoft, 1/97
ACM 97ACM 97
ACM 97
ACM 97
Top 108 reasons we predict
10. A vision and challenge
ACM 97ACM 97
ACM 97
Top 108 reasons we predict
10. A vision and challenge
7. To plan and avoid surprises– MBA schools say so– To maintain staff & budget
ACM 97ACM 97
ACM 97
Top 108 reasons we predict
10. A vision and challenge
7. To plan and avoid surprises– MBA schools say so– To maintain staff & budget
6. To extract great gobs of money– With a petaflops, we could …– It’s our “Grand Challenge” ...
ACM 97ACM 97
ACM 97
ACM 97
Top 108 reasons we predict
5. You have to: business plans, … grants
4. To stimulate new “zero growth” markets and lure investors into new ventures
ACM 97ACM 97
ACM 97
Top 108 reasons we predict
5. You have to: business plans, … grants
4. To stimulate new “zero growth” markets and lure investors into new ventures
3. A job. “Futurists” do it for pay.
2. To celebrate anniversaries and have conferences
ACM 97ACM 97
ACM 97
Top 108 reasons we predict
1. To bet on the future and earn $$$s.
ACM 97ACM 97
ACM 97
““
””
AT&T will not have AT&T will not have screwed up its screwed up its purchase of NCR by purchase of NCR by 1996.1996.
anon 1992 ($100)
ACM 97ACM 97
ACM 97
““
””
By 1997 By 1997 Video-on-Demand will Video-on-Demand will be available and be available and operating in six cities operating in six cities and...and...
Raj Reddy & Ed Lazowska1992 (winners get fed)
ACM 97ACM 97
ACM 97
““
””
By April 1, 2001 By April 1, 2001 videophones will ship in videophones will ship in 50% of the PCs 50% of the PCs and be in use.and be in use.
Gordon Bell vs Jim Gray1996 (one paper,
loser gets fed)
ACM 97ACM 97
ACM 97
ACM 97
For short-term predictions, bet against the optimist.
They are likely to be wrong. Longer term, science-based
predictions have a good chance ... wait 11 years (Mead’s Rule)
ACM 97ACM 97
ACM 97
For short-term predictions, bet against the optimist.
They are likely to be wrong. Longer term, science-based predictions have a
good chance ... wait 11 years (Mead’s Rule)
Organizations usually behave poorer than anyone can predict
ACM 97ACM 97
ACM 97
ACM 97
““
Gordon Bell “ Future of Computing” at MIT, 1972
Future computers are BOTH cheaper Future computers are BOTH cheaper AND faster model positedAND faster model posited
Semiconductor companies may make the Semiconductor companies may make the computerscomputers
We need networks, else people are the We need networks, else people are the network that serve computersnetwork that serve computers
Semiconductors (bipolar) just improve Semiconductors (bipolar) just improve for 6 more yearsfor 6 more years
““ ““ ““
ACM 97
ACM 97
Progress will continue for just 6 more years!
One generation beyond the product being worked on
If you are actually doing it as an engineer ... you tend to be very conservative
ACM 97ACM 97
ACM 97
““
”” Everything cyberizable Everything cyberizable will be in Cyberspace!will be in Cyberspace!
Gordon Bell and Jim GrayACM, 1997
ACM 97
Outline
Cyberspace: platforms, networks, and cyberization
ACM 97ACM 97
ACM 97
Outline
Cyberspace: platforms, networks, and cyberization
Some Laws from failed predictions Useful exponentials for prediction Computers we might predict
ACM 97ACM 97
ACM 97
Cyberization: interface to all bits and process information
Coupling to all information and information processors
ACM 97ACM 97
ACM 97
Cyberization: interface to all bits and process information
Coupling to all information and information processors
Pure bits Bit tokens State: places, things, and people State: physical networks
ACM 97ACM 97
ACM 97
“Everything will be in Cyberspace” Is this a challenge? Our goal?
Our quest? or Fate? Cyberization enables
new computing platforms thatrequire new networks to connect them – Infrastructure supports the content– Three evolutionary dimensions
ACM 97ACM 97
ACM 97
Region/Region/IntranetIntranet
CampusCampusHomeHome
BodyBody
WorldWorld
ContinentContinent
Cyberspace: A Network of ... Networks of ...
ACM 97ACM 97
ACM 97
DataData
Cyberspace: one, two or three networks?
TelephonyTelephony
TelevisionTelevision
ACM 97ACM 97
ACM 97
Cyberspace: A spiraling quest in 3D real space
ComputationComputation
CommunicationCommunication CyberizationCyberization
ACM 97ACM 97
ACM 97
Some more Laws from failed predictions to guide the journey into Cyberspace
ACM 97ACM 97
ACM 97
““ ””
There is no reason There is no reason anyone would want a anyone would want a computer in their home.computer in their home.
Ken OlsenPresident, Chairman and
founder of Digital, 1977
ACM 97ACM 97
ACM 97
““ ”” UNIX is Snake Oil.UNIX is Snake Oil.
Ken Olsen1987
ACM 97ACM 97
ACM 97
““
””
I wouldn’t put my I wouldn’t put my company on the company on the Internet.Internet.
Ken Olsen, ChairmanModular Computer Systems http://www.mod.com
ComputerWorld 1996
ACM 97ACM 97
ACM 97
Equating yourself to the average“user/buyer” is risky . . . unless you’re an average user.
ACM 97ACM 97
ACM 97
““
”” Navy Delphi Panel
Bernstein, 1969
Some form of voice I/O Some form of voice I/O will be in common use by 1978 will be in common use by 1978 at the latest!at the latest!
A computer will interpret A computer will interpret simple spoken sentences simple spoken sentences by 1975.by 1975.
”” ““
ACM 97ACM 97
ACM 97
““
””
Speech systems will be Speech systems will be commercializedcommercialized because because all significantall significant steps have steps have
been made.been made.
Harry OlsenRCA Labs, 1962
ACM 97ACM 97
ACM 97
Speech typewriter:
Microphone analyzer code typer pages
With translation:
Microphone analyzer code translator code typer pages
Translated speech:
Microphone analyzer code translator code synthesizer output speech ACM 97ACM 97
ACM 97
Speech predictions are optimistic and have been wrong.
Mobs and especially committees predict poorly.
ACM 97ACM 97
ACM 97
““ ””
ISDN will be ubiquitousISDN will be ubiquitous
by 1985.by 1985.
Irwin Dorros, VP Long Lines ATT, 1981
“ Unfortunately, ISDN is still likely to become the ubiquitous connection, by default, for the next 5 years.”
Gordon Bell, ACM 1997
ACM 97
Network bandwidth becomes available slower than the most conservative prediction.
ACM 97ACM 97
ACM 97
We under-estimate the devastating power ofcompanies & planners,
lawyers & government,
to foul up predictions.But to really foul up requires an econometric model!
ACM 97ACM 97
ACM 97
ACM 97
““ ””
Parallel processing Parallel processing computer architectures computer architectures will be in use by 1975. will be in use by 1975.
Navy Delphi Panel1969
ACM 97ACM 97
ACM 97
““
””
In Dec. 1995 computers In Dec. 1995 computers with 1,000 processors with 1,000 processors will do most of the will do most of the scientific processing. scientific processing.
Danny Hillis 1990 (1 paper or 1 company)
ACM 97ACM 97
ACM 97
““
”” DARPA, 1985
Strategic Computing Initiative (SCI)
A 50 X LISP machineA 50 X LISP machine
Tom Knight, Symbolics
A 1,000 node multiprocessorA 1,000 node multiprocessor
Gordon Bell, Encore
””
““
All of ~20 HPCC projects failed!ACM 97ACM 97
ACM 97
Bell Prize winners 1987-1997
Speedup: 2000X
Moore’s law: 100X
Spend more: 2X
ECL CMOS:10X
‘87 ‘89 ‘91 ‘93 ‘95 ‘97
Teraflops
100 Gigaflops
10 Gigaflops
Gigaflops
ACM 97
““ ”” Petaflops by 2010Petaflops by 2010
DOEAccelerated Strategic
Computing Initiative (ASCI)
ACM 97ACM 97
ACM 97
““ ””
1997-2010: Over a 1997-2010: Over a Petaflops is possiblePetaflops is possible
Moore’s Law 100-450x Spend more ($100M $500M) 5x Centralization of centers 3x Commoditization (competition) 3x
Gordon Bell, ACM 1997
ACM 97ACM 97
ACM 97
Parallel processing is a constant distance away.
Our vision ... is a system of millions of hosts… in a loose confederation.
Users will have the illusion of a very powerful desktop computer through which they can manipulate objects.
Grimshaw, Wulf, et al CACM Jan. 1997
““ ””
““ ””
ACM 97
Predictions are easy… especially about parallelism.
Doing is hard.
Predictions about parallelism were risky…now they are predictable.
ACM 97ACM 97
ACM 97
““
””
Vannevar Bush c1945
There will always be plenty of things to There will always be plenty of things to compute ... With millions of people doing compute ... With millions of people doing complicated things.complicated things.
memex … stores all his books, records, and memex … stores all his books, records, and communications, and ... can be consulted with communications, and ... can be consulted with speed and flexibilityspeed and flexibility
Matchbook sized, $.05 encyclopediaMatchbook sized, $.05 encyclopedia
Speech to textSpeech to text
Head mounted camera, dry photographyHead mounted camera, dry photography
”” ““
““ ““ ““
”” ””
””
ACM 97
ACM 97
Steve ManninCyberspace
ACM 97ACM 97
ACM 97
Moore’s Law is based on data and understanding.
Faith in science and a vision can be used as a predictor…
but being Vannevar Bush and being really lucky helps!
ACM 97ACM 97
ACM 97
Predicting with datawhen change is exponential
Exponentials let you predict anything you want.
But beware of how you use them!
ACM 97ACM 97
ACM 97
1. We get more
ACM 97ACM 97
ACM 97
2. New overtakes old
ACM 97ACM 97
ACM 97
3. Things get cheaper
ACM 97ACM 97
ACM 97
4. Newer & cheaper wins?
Old
NewNew
Old
NewNew
ACM 97ACM 97
ACM 97
““ ””
Bob Lucky, Vice President Bellcore, 1995
If we couldn’t predict the Web, If we couldn’t predict the Web,
what good are we?what good are we?
ACM 97ACM 97
ACM 97
Exponentials change everything
… you can’t see ‘em coming!
ACM 97ACM 97
ACM 97
Internetters growth
‘95 ‘96 ‘97 ‘98 ‘99 ‘00 ‘01 ‘02 ‘03 ‘04
Internet GrowthInternet Growthextrapolated at 98% per yearextrapolated at 98% per year
World PopulationWorld Populationextrapolated at 1.6% per yearextrapolated at 1.6% per year
12000
10000
8000
6000
4000
2000
0
ACM 97ACM 97
ACM 97
Internetters growth
InternettersInternetters
PCsPCs
TVs & PhonesTVs & Phones
World PopulationWorld Population10000
1000
100
10‘95 ‘96 ‘97 ‘98 ‘99 ‘00 ‘01 ‘02 ‘03 ‘04
““1 Gp by 20001 Gp by 2000””NegroponteNegroponte
ACM 97ACM 97
ACM 97
ACM 97
Computing and communication predictions: processing, memory, and networks
ACM 97ACM 97
ACM 97
EDSAC (c1949)
ACM 97ACM 97
ACM 97
Tera
Giga
Mega
Kilo
11947 1957 1967 1977 1987 1997 2007
Extrapolation from 1950s: 20-30% growth per year
ProcessingProcessing
ACM 97
ACM 97
Tera
Giga
Mega
Kilo
11947 1957 1967 1977 1987 1997 2007
Extrapolation from 1950s: 20-30% growth per year
StorageStorageBackboneBackbone
MemoryMemoryProcessingProcessing
Telephone ServiceTelephone Service17% / year17% / year
????
ACM 97
Microprocessor performance100 G
10 G
Giga
100 M
10 M
Mega
Kilo1970 1980 1990 2000 2010
Peak Peak Advertised Advertised
Performance Performance (PAP)(PAP)
Moore’sMoore’sLawLaw
Real AppliedReal AppliedPerformance Performance
(RAP) (RAP) 41% Growth41% Growth
ACM 97
Gains if 20, 40, & 60% / year
1.E+21
1.E+18
1.E+15
1.E+12
1.E +9
1.E+61995 2005 2015 2025 2035 2045
20%= 20%= TeraopsTeraops
40%= 40%= PetaopsPetaops
60%= 60%= ExaopsExaops
ACM 97ACM 97
ACM 97
New overtakes old
ACM 97ACM 97
ACM 97
Processor performance
1000
100
10
1
0.1
0.011970 1975 1980 1985 1990 1995 2000
RISC shiftRISC shift
CMOSCMOSmicroprocessormicroprocessor
BipolarBipolarprocessorsprocessors
VAXVAX
90009000
ACM 97
Things get cheaper
ACM 97ACM 97
ACM 97
Computers by price are stable. Lower priced computers form!
new models come at constant price and increased effectiveness
constant performance, lower price computers come
and establish new use
ACM 97ACM 97
ACM 97
Newer & cheaper wins?
Old
NewNew
Old
NewNew
ACM 97ACM 97
ACM 97
“The mainframe is dead!… and for sure this time!”
PRICE
MainframeMainframe
ServerServer
PCPC
ACM 97ACM 97
ACM 97
Predictable computers The network computer
System-on-a-chip industry
Home Area Network
Body Area Network: “on body”, “Guardian Angel”
ACM 97ACM 97
ACM 97
CMU wearable computers
ACM 97
MedtronicsImplanted
Cardioplastic
ACM 97
4 Experts Predict BionicsWired, February 1997
Hi-Fi Cochlear Implants 2005
Bionic Limbs 2013
Artificial Vision 2040
Bionic Person (unlikely)
ACM 97ACM 97
ACM 97
Observations on predicting Existence proofs are essential, otherwise
it’s faith and luck.
Numbers and data are our friends.
Bet on predictors who are grounded, intuitive, imaginative, and lucky.
Because it could, doesn’t mean it will.
It’s usually just the economics, stupid!
ACM 97ACM 97
ACM 97
Everything will be in cyberspace
ACM 97ACM 97
ACM 97
ACM 97
ACM 97
ACM 97GORDON BELL
top related