Artificial Intelligence: In our time.

Artificial Intelligence: More Real Than Ever

There was a time when the phrase "artificial intelligence" evoked images of supercomputers taking over the world, à la Skynet, the villainous computer network of The Terminator film series. Now the technology is part of everyday life. Consumers armed with smartphones and Web-connected PCs are tapping into AI software to dictate e-mail, book restaurant reservations, and find the best way to drive home. And researchers are exploring the potential for harnessing it for more important tasks and in surprising ways: One computer is playing Jeopardy! against former champions, as one step in developing computers that humans can talk to and talk with as naturally as they do each other. Here's a tour of some of the up-and-coming AI technologies.

Google Goggles

Want to know more about a book on display at the store? A movie advertised on a poster? How about the fat content of the bag of chips you just ate? Take a picture of it with your smartphone and get Google (GOOG) on the case. The Google Goggles smartphone app—as yet available only for phones running Android, although an iPhone version is expected this year—breaks pictures into searchable bits of information. That text from the movie poster, or the brand name on the bag of chips, is uploaded over a wireless network to Google, whose AI software identifies the image and retrieves related information from the Web.

Nuance Dragon Dictation 2.0

Converting the always-varying human voice into text on a computer screen is a classic task that once pushed the boundaries of computing power and artificial intelligence. We have reached the point that smartphone users can now harness the computing horsepower of the Web to eliminate the tedium of typing longish e-mails on a smartphone. Nuance’s (NUAN) Dragon Dictation for the iPhone can turn what you say on your iPhone into text you can send in an e-mail, text message, Facebook update, or Tweet. The software uploads your spoken words to servers in the Internet "cloud" and returns them to the phone as text.

Inrix Traffic Pro

What’s the drive home going to be like? If you leave now, it might take 40 minutes—but if you wait a bit, there’s a good chance the traffic will let up and cut your drive time to 20 minutes. How do you know that? Inrix Traffic Pro, a smartphone app for the iPhone and Android phones, harnesses historic and real-time data and information on such occasional things as sporting events and road-construction schedules. It uses those to predict what traffic conditions will be like. Inrix, in Kirkland, Wash., gathers 4 billion data points per month from road sensors and 3 million cars and trucks on the road using Inrix-enabled phones and navigation systems to forecast traffic on 1 million miles of road in North America.

Siri

Want to book a table for two at your favorite Italian restaurant? Tell Siri what you want—a table at a specific restaurant, or maybe a similar one in a new city—and like a trusted personal assistant, it takes care of the details. Using AI voice-recognition engines in the Internet cloud, Siri interprets what you say and gathers information accordingly. Over time it learns what you like, using details such as your location, the time of day, and your history to figure out exactly what you’re trying to say. Siri was acquired by iPhone maker Apple (AAPL) earlier this year.

IBM's Jeopardy! Challenge

In 1997 an IBM (IBM) computer named Deep Blue defeated chess champion Garry Kasparov in a six-game chess match, proving that computers could beat experienced players at that game. But what about the TV game show Jeopardy? The game-show host’s questions, always phrased as answers with puns and in-jokes that are easy for a human to understand, are trickier for computers. IBM has built a supercomputer it calls Watson that can compete with human champions from the TV show at a Jeopardy-like set IBM created in Hawthorne, N.Y. Eventually IBM hopes the work will lead to computers that can respond to complex questions from humans asked in everyday speech and give the right answers.

Soccer-Playing Robots

First chess, then Jeopardy. How about the most popular sport in the world? Claude Sammut, a professor of computer science and engineering at Australia’s University of New South Wales, predicts in a paper published in the academic journal Cognitive Science that a team of robots will have both the physical and reasoning abilities to best a team of human soccer players by the year 2050. It’s no small task, Sammut writes: "To achieve this, or come even close, the robots will have to be able to sense and act in completely unstructured environments. This will require major advances in perception, decision making, learning, and cooperative behaviors." Researchers have been conducting the RoboCup—a international robotic soccer competition—since 1997, and the current generation of players are midget-size Aldebaran Nao robots. As yet there’s no threat to humanity’s control of the World Cup.

Expertmaker

Expertmaker, a Swedish startup, demonstrated a set of tools it says will make it easy for non-programmers to harness artificial intelligence to build subject-specific Web search engines that can be used in business or by consumers. It demonstrated one tool geared to helping players choose from a range of new video games based on the games' intensity, complexity, and how well they are reviewed. The technology is also behind a smartphone app called Pasta Explorer that helps consumers discover pasta recipes based on personal preferences. The company plans to license its technology to companies that will use it to underpin apps and services that deliver tailored search results.

Memristor

Here's something humans can do that computers can’t: remember. Turn a computer off, and it forgets everything it was doing only seconds before. (Of course, it has memory storage devices to help ….) A memristor—a mashup of "memory" and "resistor"—is an electronic circuit that remembers its state after the power goes off. This is similar to the way a synapse works in the human brain. First theorized in 1971, researchers at the University of Michigan are now using memristors to build a computer that mimics the brain of a cat. Professor Wei Lu says the simulators now require 140,000 CPU chips, and even with that computing power a cat can still do simple cognitive tasks, such as face recognition, 83 times faster. Still, Lu says memristors will in time make possible a computerized cat brain that’s no bigger than a two-liter soda bottle.

Context-Aware Computing

Today computers are great at retrieving information that may—or may not—suit your needs quickly. (That's why Google is as popular as it is.) But what if they could make suggestions as if they were a good friend or relative who knows you well? Chipmaker Intel (INTC) is working on what it calls context-aware computing, whose devices know what restaurants you like and don’t like. They suggest relevant choices based not only on your current location but also on your historical tendencies, your schedule, the weather outside, and scores of other factors. In theory they’ll eventually anticipate your needs and give helpful, unsolicited advice. For example, knowing you have an important business meeting today, your smartphone may suggest you leave a littler earlier than usual because traffic on the way is tied up.

Brain-Controlled Wheelchair

What if the human brain and artificial intelligence were to team up? For starters, they might allow a person in a wheelchair to get around more easily. Researchers at Switzerland’s École Polytechnique Fédérale de Lausanne have developed a thought-control system for navigating in a wheelchair. Think of moving left, and the chair moves left. Think forward, it moves forward. AI is used to make it easier for the controller. Two cameras connected to image-processing software help navigate around obstacles. In the future, the system will learn how to tell the difference between obstacles and objects the user may want to approach. For instance, it might recognize a checkout counter in a store and know to pull closer.

Intel's "Mental Typewriter"

Researchers at Intel are developing technology that could one day be built into a hat that would determine what word you’re thinking of at any given moment. The point: Why type a search query when you can just think it? A room-size prototype can already determine when a test subject is thinking about such things as celery and airplanes. Tinfoil hats, anyone?