how do industries evolve - part 2
Five Myths of Technology Change
A/Prof Jeffrey FunkDivision of Engineering and Technology
ManagementNational University of SingaporeFor information on other
technologies: see http://www.slideshare.net/Funk98/presentations or
Exponential Change: What drives it? What does it tell us about the
future?
http://www.amazon.com/Exponential-Change-drives-about-future-ebook/dp/B00HPSAYEM/ref=sr_1_1?s=digital-text&ie=UTF8&qid=1399871060&sr=1-1&keywords=exponential+change
OutlineMyth #1: Performance vs. time curves resemble an
S-curveMyth #2: Slowing rate of improvement in old technology
drives development of new technologyMyth #3: Product design changes
drives performance increases and process design changes drives cost
reductions, with product preceding process design changes in life
cycle Myth #4: Costs fall as cumulative production rises in
learning curveMyth #5: All technologies have the potential for
rapid rates of improvementsPulling these myths and realities
together
TimePerformanceMyth: New Technologies Follow S-Curves
Emergence of New Technology
The Theory for Purported S-CurvesImprovements accelerate as
research funds moved from old to new technology in responseto
increases in demand for new technology or to slowdown in rate of
improvement in old technology (Foster, 1986; Garcia and Clantone,
2002; Utterback, 1994)Acceleration may also occur as technology is
better understood by scientists and firms, constraints are
overcome, and complementary technologies developed and implemented
(Butler, 1988)For later part of purported S-curve, rates of
improvement slow as cost of marginal improvements increases and
natural limits emergeresearch funds then move to still newer
technology and thus acceleration in newer technologys rate of
improvement (Foster, 1986; Butler, 1988; Utterback, 1994)
S-Curves make it easy to fall for hype
Tell my story5
Lets look at some real dataMostly straight lines on a
logarithmic plotbut with some deviations
Limit?Slowdown?
Slowdown?
Slowdown?
Acceleration?Acceleration?Acceleration?
Acceleration
Acceleration?
No Evidence for an S-CurveNone of the 32 time-series curves
display classical S-curveSecond half of S-curve, i.e., limits, only
evident in one technology, best laboratory efficiency of amorphous
silicon solar cells (Figure 1.c)However, output (kwHours) per
dollar is probably still risingsimilar to output per dollar for
crystalline silicon solar cells (See Figure 1.c.). Another data
base (Economist, 2012) shows continued reductions in cost beyond
2003 (Nemet, 2005), which is last data point Figure 1.c
No Evidence for an S-Curve (2)First half of S-curve, i.e.,
acceleration, only evident in one technology, cellular
telecommunications (Figure 1.u)This acceleration is expected since
cellular phones were first used for voice communicationdata speeds
only became important in late 1990s as displays reached levels
necessary for data speeds to become important; this explanation is
consistent with the theory of S-curves (Foster, 1986; Butler, 1988;
Utterback, 1994)Better measure of performance for early years of
cellular phones would probably be number of voice conversations
possible per unit of spectrumWe will show this data later this
semester. Not enough data points, but straight line
No Evidence for an S-Curve (3)Several other curves deviate from
straight lineSlowdowns Mobility of organic transistorsComputer
tomographyBut this doesnt mean limits will soon be reached since
other technologies have seen slowdowns followed by
accelerationsMagnetic recording density of tape and disks (due to
introduction of giant magneto resistance)Number of sequenced DNA
base pairs - due to introduction of new technology (454,
Illumina
RealityMost curves more closely resemble a straight line on a
logarithmic plot than an S-curveWhat does a straight line on a
logarithmic plot mean?Lets look at the statistical analysisLinear
modelLogarithmic modelLogarithmic model with time squared
Regression Analysis of Performance and Cost vs. Time for Various
Technologies
Regression Analysis of Performance and Cost vs. Time for Various
Technologies(continued)
Why arent there Early Accelerations? Improvements build from
past improvements and the extensions to the knowledge base that
these improvements bringThus, rates of improvement are relatively
constant over timeR&D is decentralized and it has become even
more decentralized over last 50 yearsmost researchers create their
own research plans and try to publish something new and
differentfunding for new technologies is also highly
decentralizedThus, research efforts quickly move to new
technologies as research results are presented and published and
accelerations dont occur
18
OutlineMyth #1: Performance vs. time curves resemble an
S-curveMyth #2: Slowing rate of improvement in old technology
drives development of new technologyMyth #3: Product design changes
drives performance increases and process design changes drives cost
reductions, with product preceding process design changes in life
cycle Myth #4: Costs fall as cumulative production rises in
learning curveMyth #5: All technologies have the potential for
rapid rates of improvementsPulling these myths and realities
together
Myth and RealityMythSlowdown in rate of improvement for old
technology causes new technology to be developed and
improvedRealityNew technologies are being developed long before the
rate of improvement slows in the old technologyOften multiple
technologies are being simultaneously developed as replacements to
an old technologyThe following slides show examples of this
Rates of Improvements for Technologies that Might be Considered
Substitutes
Did Slowdown in Fluorescent Lighting Lead to Improvements in
LEDs/SSL?(SSL = Solid State Lighting)Source: Lima Azevedo, Granger
Morgan, Fritz Morgan, The Transition to Solid-State Lighting,
Proceedings of the IEEE 97(3)Luminosity per Watt
Semiconductor LEDs now have higher luminosity per watt than do
incandescent lights and will soon pass those of fluorescent
lights26
Luminosity Per Watt
Lets Look at a Statistical AnalysisAre rates of improvement for
the old technology slower after the new technology is
introduced?Many ways to do this testWe could introduce a lag, say
of 5 yearsLets just contrast rates of improvement before and after
performance metric is recorded for the new technologyTest for
change in the slope
TechnologyDomainNewTechnologyOld Tech-nologyRates of Improvement
for OldTechnologyNumber of Data PointsP-ValueBefore NewAfter
newBefore AfterLightingLEDsFluor-escent1950-1970: 0.65%1970-2002:
0.37%102.0012OLEDsLEDs1968-1986: 38.7%1986-2008:
37.0%515.99DisplaysQuantum DotsOLEDs1987-1996: 35% 1996-2011:
22%38.90ElectricityCrystalline Silicon SolarFossil Fuel1882-1957:
5.4%1957-1972: 2.7%144
