October, 2003
Personal connectivity is a stew made of social, technical and political
ingredients. In recent years, technology has been more of an enabler than
a definer of things people do. SMS was transformed into a daily service
by users, flash mobs were never envisioned by the inventors. There is
little reason for this trend of social invention to abate.
A good guess is that renegade systems will persist. The trick to scale
a system is to minimize the infrastructure required for it. 802.11 has
a sparse and random infrastructure and it has been adopted widely. 3G
has a comaparatively huge installation and scaling problem -- the investment
needs to be made up front.
Therefore, a good area for development is in systems that have the viral
nature of 802.11, the scalability of fax machines, and the social freedom
of napster. I think this can be done with intelligent radios that act
cooperatively to build networks on-the-fly in dense areas, but you can
pick your favorite in your own industry. In the short term, we have not
yet used up the potential for mesh 802.11 and we are not likely to unless
some magical, mobile broadband service takes hold. Given that there are
no good ideas for such a system, we can remain at the 802.11 bandwidth
plateau for quite a while.Broadband to the home
The trick in building an infrastructure is to define it to be minimal,
long-lived, and generic. The US National Defense and Highway act did a
good job, Minitel did not. In both the US and abroad, we will need infrastructures
for communications that can last for ten years and support the widest
possible uses.
Fiber to the home is not interesting.
The models for penetration are based on broadcast, not symmetric access,
and are outdated.
A key issue in radio broadband will be immunity to attack. Currently,
hackers vandalize the internet regularly but they do not go after broadcast
and telephone services. (They could) Perhaps the best response is to provide
another forum for hacking, such as the financial system.
Broadband wireless.
This is an area where linear thinking is retrograde. There are too many
revolutions left in radio for simple extrapolation. The driving force
is software radios, not in the limited sense that the word is currently
used but through fully programmable RF systems. Once the radio becomes
digital, then the rules of 80 years of spectrum thinking are reversed:
we optimize for transmission not for cheap receivers. Further, a programmable
radio can do a complicated task almosst as easily as a simple one.
Remember Lippmans's first law: Given the choice between and expensive
and a cheap technology, always chose the latter. It will be cheap enough
when it catches on, and the flexibility is worth the advance bet.
Most important, it moves the radio intelligence to the leaves and opens
communications up to those who never had a seat at the table. It can become
an embedded industry.
The magic in the human interface is to watch the kids. You are welcome
to track speech technology, but the driver is not the technology, it is
the users. We have had sufficiently good recognition and synthesis for
years. Incremental improvements are not the issue nor is some magical
threshold in ease-of use.
Twenty-five years ago, we build a recognition-based phone answering machine.
At the time, it was a successful demonstration, but answering machines
were infamita. Now, ones that are much worse are common. Why not learn
from that experience?
(Lippman's second law: Let your customers be the inventors. This is "centralized virality." The worst option is a bunch of fat fourty-year-olds deciding what people will buy. Even if only 1 in 100,000 PC user program them, the sheer numbers overwhelms what IBM or Oracle can hire. Where do you think the innovation will come from?)
(Lippman's third law: Never teach anything to anyone you love. Let someone else do it. How many couples have shot each other over bridge bids or divorced after tennis "lessons?")
Automotive:
If you can build a "wireless car", one with no copper outside
of the engine compartment, you would have a real winner. The interference
and reliability constraints on the auto industry delay innovation and
cost in power and weight.
Similarly, an open architecture car opens the domain for people to install
new stuff regularly. That seems like a good idea.
Intelligent highways should come from intelligent cars, not the infrastructure.
WHy not make rush hour an adhoc network?
Internet Enablers:
No one seems to have fully gotten the message that the internet is a great
equalizer. E-Bay tells is that everyone is a merchant; napster tells us
that everyone is a distributor; MP3 tells us that there are more artists
than you thought. Still, we build networks that replicate CBS and NBC,
thinking that is where the money is.
The next threshold could well be education. Suppose we decentralized it
entirely and said that everyone was a teacher?
Multimedia:
For thirty years, the missing element in multimedia was imagination. Engineers
seem to think that it is fidelity and have gone so far as to debug VR,
but people can use multimedia in crude forms if it is worth it. I would
leave this entire topic to the audience and be a follower. Develop basic
technologies and avoid considering applications.
