Background:
Early in 1992, I
helped prepare a 5-Year Plan for the Huntsville Research
Operations of Georgia Tech. This compares the forecasts I made
with what actually happened. The text in blue italics is what I
wrote in 1992, and the black text below it is the year 2000
update.
A One-Time Quantum Leap in Electronics Technology Over the Next Few Years
1992:
"As
one who has made a hobby out of forecasting progress in
electronics technology , I foresee an unprecedented one-time
surge in electronics progress over the next few years. I think
this growth spurt began in 1991 and will continue through 1995,
and perhaps into the next century. It appears to me that several
"waves" are converging at once.
"First, telephone bandwidths should widen dramatically for
the first time, even as data compression techniques permit
reductions in bandwidth requirements."
Year 2000
Update:
Well, yeah, but
bandwidth improvements didn't really begin to show up until 1999,
and it will be the early years of the 21st century before most
households enjoy the benefits of wide band service.
In 1992, the
telcos were touting 128-kilobaud ISDN service for $30 to
$40 a month.. But when ISDN actually arrived in 1995, the telcos
charged $70 to $100 a month for 64-kilobaud service,
pricing it out of the market. On the other hand, DSL
services run 1,000 to 1,500-kilobaud downstream for as low as $40
a month, and provide far more capability than ISDN ever promised.
1992:
"Second, cellular systems should advance rapidly with the
advent of micro-cellular networks, wireless computer networks,
and direct satellite relays,".
Year 2000
Update:
That has more or less
happened, with the proviso that the Iridium satellite service was
a bust. And "you ain't seen nothin' yet". A transition
from a wired world to a wireless world appears, in my opinion, to
be in the offing between 2000 and 2010.
"Third, IBM and
Apple's announcement of the PowerPC project, which is scheduled
to lead to 30-fold increases in high-end PC speeds by 1994, has
forced competitors to accelerate their plans, promising desktop
computers by 1995 which rival the supercomputers of 1990 . For
example, Intel has advanced their introduction of the 80586
(scheduled for introduction this fall) by at least a year, and of
the 80686 (now scheduled for introduction next year) by at least
two years. Not to be outdone, TI and Sun Microsystems have
announced plans for a 600-MIPS "SuperSparc" to be
marketed in 1993. At the same time, supercomputers are projected
to accelerate from speeds of 16 gigops for the Cray Y-MP to a
1,000 gigops by or before 1997, and perhaps, to multi-terops
speeds by the year 2000. Again, this is a one-time jump in speeds
as supercomputers shift from largely serial operation to
massively parallel processing."
In 1992, a
vice-president of Hewlett-Packard had announced that their new
RISC technology would lead to 500-MIPS HP microprocessors by
1995. I predicated the above forecast (that 1995 computer speeds
would be about 30 times as fast as1992 computer speeds of,
perhaps, 15-16 MIPS for a 40-megahertz 486) upon this and upon
similar announcements from Apple and IBM. Normally, one might
expect about a 4-fold increase in computer speeds over a 3-year
period, so a 30-fold quantum leap would have been rather
profound. In practice, computer technology didn't quite make it.
The Intel Pentium Pro, released in December, 1995, clocked out at
about 328 MIPS, or about 20 times the speed of a 1992 486. Still,
it was an impressive delta. Interestingly enough, it was Intel
with a CISC (Complete Instruction Set Computer) chip that won the
race, rather than the RISC family of chips.
The good news is that
computer technology is still surging ahead, with multi-gigops
chips due before year's end (the 1.4 GHz Willamette, and the
Itanium chip).
Intel delivered its
terops Touchstone Delta processor to DARPA on schedule in 1997. I
believe that a 100-terops computer is in the works for
2005.
"Fourth, digital
signal processors (DSPs) are scheduled to appear on Macintoshes
and PCs toward the end of this year. (Fixed-function graphics
accelerators are already built into our high-end graphics cards,
but programmable DSPs are becoming available.) To give an example
of what this means, the new AT&T AVP-1000 DSP chip set
operates at speeds up to 25 gigops! These DSPs achieve their
blazing speeds through parallelism (by performing the same
operation of 32 to 64 parallel data streams), hard-wired
functions optimized for specific purposes, and 8 to 24-bit data
paths. In other words, they are a poor man's parallel processor.
Speed increases this large may portend qualitative changes in the
way we use computers. For example, voice recognition, simple
household robotics, and desktop video teleconferencing are three
of the applications which might be I think might be expected to
surface within the next three years."
It's now 8 years later,
and voice recognition, while available, is crude, and designed to
run on older computers, with no DSP support. However, speech
recognition has come into commercial use in conjunction with
various voice-response systems. Speaker-independent,
continuous-speech voice recognition systems are apparently
available but they didn't surface by 1995. And household robotics
and desktop videoconferencing certainly didn't surface by
1995. On the other hand, commercial robotics application have
continued in the background, including such inventions as the
automatic car wash, the ATM, anti-lock braking systems, and, most
recently, automated trash pickup. This last item deserves a
little embellishment.
About a month ago,
Huntsville transitioned to automated trash pickup, reducing the
number of men on a trash truck from 3 (the truck driver, plus two
trash handlers) to 1 (the truck driver). The new robotic trash
trucks have a pair of hydraulically-operated arms that sense the
(special) trash can, reach out and girdle it, lift it high
overhead, and shake its contents into the yawning maw of the
truck's cylindrical tank. These trash-collectors are robotic in
every sense of the word, and yet, there's nothing revolutionary
about them. In commercial settings, robotics has continued its
penetration into the workplace. Voice response systems, smart
search engines, improved character recognition software, and
web-'bots are examples of AI in the consumer marketplace. The
diffusion of robotics and AI is slow enough and subtle enough
that we take it in stride. On the other hand, anthropomorphic
household robots are still far away, although automatic,
visually-guided carpet sweepers and lawn mowers might appear in a
few more years.
Desktop
videoconferencing is, presumably, hostage to the widespread use
of wideband telecommunications systems. However, in another three
or four years, it may affect distance learning programs.
"It may be that PC speed enhancements will track
supercomputer speed increases beyond 1994, through the adoption
of parallelism and the addition of programmable digital signal
processors."
The digital signal
processors are there in the form of disk controllers, and
graphics and sound cards. However, they're probably not
programmable, although 30-gigops, multi-gigaflop, programmable
DSP's are now available for $100.
So far, parallel
processing hasn't been necessary to speed up desktop computers.
Clock speeds have gone from, perhaps, 15 MHz in 1990 to 1.5 GHz
by the end of 2000. IBM has promised 4.5 GHz speeds by 2003, and
SiGe chips might be capable of 3 to 5 times that speed. 10 GHz
chips have been predicted later in the decade.
"Fifth, both voice
and video (though not broadcast video) may be expected to go
digital within the next few years. This will permit such
techniques as time-averaged noise filtering, enhanced definition
TV using interpolation combined with image enhancement, and
model-based encoding. I suspect that DoD may adopt digital
communications techniques as they become commercially available."
This is a slow
and continuing revolution. We're in the long transition
period during which both digital and analog transmission schemes
exist. Conventional telephone lines still operate in analog
mode from the counter to the curb, and are digital beyond that
point. Internet based telephony is all-digital. VCR's are analog
devices; DVD's are digital implements. Eight years later,
audio recorders and camcorders are switching from analog to
digital.
"Sixth, HDTV
apparatus should enter the marketplace, providing higher quality,
higher resolution video capture, storage, transmission and
presentation equipment for consumer and military
applications."
HDTV was delayed
several years over compatibility squabbles, and now,
although it exists, it's certainly not yet mainstream. I would
expect a few more years to pass before HDTV displays,
camcorders, and players are common and competitively
priced. However, digital still cameras are preparing to displace
most film cameras.
"Seventh, large
high resolution, flat-screen, color TV displays should become
affordable, permitting practical stereo TV. Truly lightweight,
head-mounted displays driven by RF links may become feasible."
Large, flat-screen
color TV displays have really been a Lorelei. LCD color displays
never became either large or cheap. Stereo (3-D) displays haven't
happened, either. The same kinds of displays that were common in
1992 (CRT's and CRT projection displays) are common today. Plasma
panel and CRT projection displays are the current front-runners
for HDTV output. Truly lightweight, head-mounted displays driven
by RF links exist. They're a little expensive, and so far, they
haven't really caught on. Another thing that hasn't really
caught on is virtual reality, except in support of computer
games.
"Some of these "waves" have been developing for several decades, and happen to be converging at about the same time."
Summing it up,
I was too optimistic about how fast some of these
improvements would occur. The original forecast had an aura of
excitement about it, in keeping with its role as a wake-up call
for an impending technological tour-de-force, but developments
have been more evolutionary than revolutionary. Telephone
bandwidths have widened slowly, and for most people, they haven't
widened at all. Bandwidth compression schemes have no doubt
improved mightily since 1992, and by now, they can provide
cameo-screen, herky-jerky video over voice-grade lines.
Coupled with wide-band service, as it penetrates the lay of
the land, high-quality audio and video should be widely available
by 2005, but it certainly wasn't available in 1995. Cellular
phone service has arrived as advertised, tying into fiber-optic
cables. Computer speeds didn't quite improve by a factor of 30 in
3 years, but they didn't miss by much, and they certainly
outstripped the 4-to-1 acceleration that would normally have been
expected. Speaker-dependent continuous-speech recognition
software packages are available at your local computer store for
as little as $49.95, including a noise-cancelling microphone, but
they aren't yet ready for prime time. Household robots are still
over the horizon, although robotics has continued to diffuse into
industrial applications, such as voice response systems. There is always this time lag in commercial settingsbetween what can technically be accomplished, and
what is cheap and in majority use.
At the same time, there
were some wonderful and revolutionary developments I missed:
(1) The Internet
(2) The sudden acceleration,
beginning in 1995, of magnetic disk progress.
(3) The dramatic improvements and
price declines in ink-jet printers.
(4) The dramatic price declines of
scanners.
(5) The continuing validity of
Moore's Law beyond 2000, when it was supposed to hit a
"brick wall"
(6) Computer prices dropping into
the few-hundred-dollar range.
(7) The Japanese eclipse, and the
resurgence of the U. S. and Europe as world technological
leaders.
In short, the
developments have been as dramatic as what I forecast, but not
for the reasons I expected. The market-driven events have been
delayed several years, while the rise of the Internet has helped
offset these other delays. (I include practical speech dictation
as a lagging, market-driven technology. Household robots, though,
are another matter.)
The following
is an additional set of notes that I also produced in 1992,
together with their updates.
Technological Opportunities
It appears to us that there a number of technological waves which are about to converge to produce an seldom-precedented surge of electronic progress.
(1)
Telecommunications is undergoing revolutionary changes as:
- bandwidths
increase from kilobaud rates to tens of kilobaud rates to
(putatively) megabaud and later, gigabaud rates;
Bandwidths have
increased from multi-kilobaud rates to 33.6 kilobaud upstream,
and about 50 kilobaud downstream. For me personally, running ADSL
(since 1999), bandwidths have reached the megabaud rate
downstream and the 400 kilobaud rate upstream. After years of
somnolence, a telecommunications revolution is finally in full
swing.
- both audio and video
communications protocols switch from analog to digital mode;
The conversion
from analog to digital is taking place gradually. By 2006, the
FCC has mandated that all TV transmissions are to be digital.
- digital data
compression algorithms reduce bandwidth requirements
Next year,
streaming video, using the latest data compression algorithms, is
supposed to go mainstream.
- it becomes a
"wired world"
The "wired
world" has certainly become a reality. The next drift will
be toward a wireless world.
- Programmable
digital signal processors and application-specific integrated
circuits approach image processing speeds at which machine vision
systems can rival human vision.
That may already
be feasible but it hasn't taken hold.
- Programmable
DSPs become fast enough and high-speed RAM becomes cheap enough
that large-vocabulary, speaker-independent, voice recognition
subsystems capable of recognizing continuous speech are feasible
and inexpensive.
Large vocabulary,
speaker dependent voice recognition software is cheaply
available, but I'm not sure just how capable hardware-augmented
speech recognition is in commercial settings. Speech recognition
add-in boards don't seem to have caught on for home use.
- RISC-based
chips approach speeds at which autonomous, free-standing robot
devices capable of performing industrial tasks become
several-orders-of-magnitude more feasible than they are today;
Two-orders-of-magnitude chips are available, but autonomous,
free-standing robotic devices aren't exactly ubiquitous.
"RISC" is no longer a buzzword. I'm anticipating the
first visually-guided automatic carpet sweepers and lawn mowers
no earlier than 2005, and then only as curiosities. By 2010, they
could possibly be mainstream, and perhaps by or before 2015, they
might be on sale at Walmart.
-
Fractal-based, mosaic screen-projection videophones (if
necessary, using multiple cameras, multiple phone lines) Need to
rework a videophone so that its display can be used as a
projection panel. Need to develop a high-resolution videophone,
perhaps using multiple lines to transfer data.
Sounds like some
good ideas. However, fractals are out. Evidently, there are some
problems with using LCD panels for screen projection. And by now
wide-band lines can give us adequate videophone capabilities,
with any kind of display and speaker system for output.
- Remote
teaching of courses over videophones (the extended classroom)
Now known as
"distance learning", and shows great promise.
-
Computer-aided instruction at the undergraduate and graduate
level
Some coursework
is available.
- Conference
call capability for a large number of callers (needed to support
the remote teaching of courses)
It's known as
NetMeeting, and it's available from Microsoft and Netscape.
-
Object-oriented encoding/local animation, based upon machine
vision.
It's known as
MP4, and it's
hot!
- Robotics
kits
Lego Mindstorms
- The
automated office: groupware, two-way video intercoms,
at-your-desk teleconferencing, the virtual office, online
document storage and retrieval, automatic font construction,
Not as much in evidence
as I would have expected by now. Videoconferencing is probably
key.
- Online
document storage and retrieval system.
Available in
spades.
- Text, font,
format compression
Part of our
infrastructure by now.
- Industrial
automowers, industrial carpet sweeping, cleaning robots,
"ScoutAbout"
"Scout
About" never made it to market. One automower is on the
market. No visually-navigated robots are in widespread use. Might
see experimental prototypes in 2005.
- Machine
vision.
Used for visual
inspection in manufacturing
- Robotics
Special purpose
"robots" are gradually infiltrating the world.
- Cellular
wireless local area networks tied into ISDN and later, to SONET
networks
They're tied
directly to fiber-optic trunks.
- Miniature
lunar and Martian solar-powered mapping/mineral prospecting
robots (tested first on the ground) (self-repairing?
mutually-repairing?)
NASA is
pioneering in this.
- Application
of the above to terrestial mapping, prospecting.
Carnegie Mellon
robots are being used for Antarctic missions.
- A
battery-powered, solar-rechargeable Martian UAV
- Definition of
the minimum equipment which must be transported to Mars in order
to construct a micro-industrial base.
I don't know
their status.