lunes, 22 de octubre de 2018

Ficha del recurso:


Vínculo original en Bussinesweek
Olga Kharif

Fecha de publicación:

martes, 21 de septiembre de 2010

Última actualización:

martes, 21 de septiembre de 2010

Entrada en el observatorio:

martes, 21 de septiembre de 2010



Archivado en:

Artificial Intelligence Goes Mobile

Some of technology's biggest companies are developing AI-related hardware, software, and components to run on smartphones and tablets. Might a wireless device rebalance your investment portfolio?

Artificial intelligence is going mobile. The technology that can help machines behave more intelligently, popularized by such films as 2001: A Space Odyssey, is finding its way onto tablet-style computers and other handheld devices.

Researchers at International Business Machines (IBM) have created a machine called Watson that can sift through a terabyte of data and crank out answers to complicated questions in three to five seconds. A version of the software that runs Watson could reside on a doctor's tablet computer in three to five years, analyzing test results to proffer a diagnosis, says Dave Ferrucci, a senior manager at IBM. Or it might analyze real-time market data and recommend ways to rebalance an investment portfolio—from a smartphone, he explains.

"We are right at the dawn of a new age of the capabilities of machines," says Geordie Rose, founder of D-Wave Systems, which develops chips for computers that run artificial intelligence applications.

Other companies working on artificial intelligence-related software, components, or hardware include Intel (INTC), Hewlett-Packard (HPQ), Google (GOOG), Apple (AAPL), AT&T (T), and Sprint Nextel (S).

mightier processors, faster networks

At stake is a large and quickly growing market. Global revenue from mobile speech recognition—one of myriad artificial intelligence applications—may rise to $780.5 million in 2015, from $160.3 million this year, according to ABI Research in Oyster Bay, N.Y.

Artificial intelligence is currently used in such areas as defense, gaming, and finance. Soldiers can use it to interpret what is seen by remotely controlled robots. AI software can interact with human players to make video game play more challenging. And banks have long used it to detect transactions that fall outside the norm—for instance, to identify fraud.

The technology is headed for mobile devices as smartphones become more prevalent and capable of handling complex tasks. More than half of Americans will own a smartphone by the end of 2011, according to market researcher the Nielsen Co. Phones and mobile tablets sport more powerful processors that let users run complex artificial intelligence apps on them. Faster wireless networks also help deliver AI applications to mobile devices.

"We are trying to reach that Star Trek dream," says Mazin Gilbert, executive director of technical research at AT&T Labs. "We spent decades investing in this technology. If you can put AI with mobility, it really—significantly—expands the number of applications and services" you can provide, he says.

Intel's wireless mind-reading hat

AT&T has devoted more than 1 million research hours to develop AI technology that can convert speech to text, engage in dialogue, and conduct searches in response to spoken questions. This year, AT&T will let developers create mobile apps that tap into its artificial intelligence engine.

AT&T is also exploring ways to let people use voice commands to get directions while driving and control appliances such as television sets. "We have 50 to 60 cool things [in the lab] no one's seen yet," Gilbert says.

Some advanced applications of AI will take time. Intel, the world's largest chipmaker, is developing a technology that could be built into a wireless hat and work as a "mental typewriter." By scanning brain waves, it can guess what a person is thinking and predict search queries. A prototype of the software can recognize about 60 words, such as "airplane" and "celery," says Dan Pomerleau, a researcher at Intel Labs in Pittsburgh, Pa. It could take more than 10 years for its vocabulary and accuracy to improve enough for real-life uses, he says.

Other challenges abound. AI software still struggles to identify objects visually. Recognizing a cat, for example, can be difficult. The animal can stretch or curl up and its various shapes and sizes can confuse a machine.

HP aims for the cloud

To give its machine more brainpower, Hewlett-Packard is developing an analog computer that replicates the workings of a human brain and can crunch data many times faster than current machines.

"We want to supply the major part of hardware for the cloud and cloud services," says Stan Williams, founding director of a lab at HP Labs.

AI technology can be particularly useful in sorting through the ever-rising tide of digital data. In just six years, social site Facebook has gained more than 500 million active users, many of whom post photos and comments daily. Google answers more than 1 billion search queries a day. Every minute, 24 hours of video is uploaded to Google's

Nuance Communications (NUAN) is working on software that mines people's interests, likes, and dislikes, as reported on Facebook and microblogging site Twitter, to provide more relevant results to search queries. Says Gary Clayton, Nuance's chief creative officer: "It's all about having all this data and being able to predict the needs and wants of an individual."