A Mixed Bag for U.S. Institutions
U.S. academic and nonprofit institutions and technology investment firms that responded to two recent surveys saw their technology licensing revenue grow 7.8 percent between 2002 and 2003, according to the Association of University Technology Managers. The number of patents issued to these institutions also increased, by about 10 percent, but the number of startups they formed dropped from 398 in 2002 to 364 in 2003. Data on all survey respondents for the past 11 years – including irregular participants – show that the number of U.S. patents issued each year, as a proportion of applications filed, is also declining. New York institutions receive the most licensing income, while those in California lead the field in R&D spending, patenting, and startup formation.
Spies Like Us
Search engines find keywords in mountains of information. But suppose you need – really need – to look for what data miners call “roles and relationships”: who’s doing what, with whom, where, when, and how? For answers, the CIA is turning to a Palo Alto, CA-based startup called Attensity whose linguistics-based software extracts that kind of information from e-mail, electronic message boards, and other free-form text. David Bean is Attensity’s cofounder and CTO.
How do linguistics geeks end up doing spook work?
A couple of months after September 11, the CIA’s venture capital office, In-Q-Tel, called us and said, “We hear you can extract roles and relationships from unstructured text.” We said, “Yes, we’re working on it.” They ended up as our lead investor.
You’ve also got Whirlpool and General Motors as clients. What’s in it for them?
We can deal with input that is noisy – run-on sentences, fragments, misspellings. That’s what you typically get in the real world of e-mail and product support. We can comb through it all and pinpoint where customers are having problems.
So have you all got top-secret clearances?
Actually not. Part of In-Q-Tel’s purpose is to work with companies that are not typical players with the government. We give them the software and teach them how to use it, and then they go and do things they don’t tell us about.
R Deficit = D Inflation
The cost of developing new weapons systems is on the rise, and the U.S. Government Accountability Office (GAO) blames a too early emphasis on product development. The GAO evaluated 54 defense projects and found that just 15 percent had a sufficient scientific basis when they were initiated. The GAO’s estimates of the R&D costs of 26 projects (those that had enough financial data available to analyze) were 42 percent higher than initial estimates – an increase from 21 percent to 26 percent of total costs.
First-in-Class Ain’t What It Used to Be
In the pharmaceutical sector, the amount of time in which a new product has a market all to itself has diminished considerably over the past several decades. By the late 1990s, it wasn’t uncommon for a first-in-class drug (the first drug to use a particular molecular mechanism to treat a condition) to have several competitors in various stages of clinical development – even before it was approved. A few decades ago, a new drug would have a corner on the market for about eight years. Now, that figure is closer to two years.
Proclaiming Rain Falls Mainly to a Plane
Since january, newly developed sensors affixed to 64 commuter planes owned by Eagan, MN-based Mesaba Airlines (a Northwest Airlink affiliate) have been sending real-time data on humidity, temperature, wind speed, turbulence, atmospheric pressure, and location to a central station on the ground. In early findings, data from the planes is allowing forecasters to predict with far greater precision the arrival time of precipitation or freezing temperatures and the likelihood of severe thunderstorms or fog.
Readings from the sensors fill a huge gap in the data meteorologists collect. The lower ranges of the atmosphere, below 6,000 meters, are where weather forms. But currently, in the entire United States, 69 weather balloons take just two daily “soundings” in the lower atmosphere. In the new project, each commuter plane takes measurements every time it takes off and lands – which adds up to 600 to 800 total soundings daily. The planes also send regular dispatches from cruising altitudes.
“The amount of data we’re getting is just incredible,” says Jeff Last, a Green Bay, WI, meteorologist with the National Weather Service, one of the agencies involved in the project. In one test this winter, a forecast using data from the sensors accurately predicted a snowstorm’s arrival, while a traditional forecast was off by three hours.
In addition to the National Weather Service, the project involves other government agencies, including NASA, and several universities and private companies. AirDat of Morrisville, NC, the company that processes the sensor data, hopes to eventually sell climate information to airlines, agricultural businesses, and others interested in a sharper weather picture.
From “Visual Aids”
(July 1955, p. 468)
Become an MIT Technology Review Insider for in-depth analysis and unparalleled perspective.Subscribe today