In the next decade, the United States will use increasingly capable artificial intelligence (AI) to greatly reduce the cost of health care, accelerate research and development into new medicines, improve cars and roads to reduce gridlock, and even regain much of the manufacturing base we lost to countries like China, say researchers in computer science, robotics, and management. They claim that AI will soon change the work of doctors, nurses and teachers across the country, create entirely new businesses, and radically remake industries already in existence.
AI is already directing traffic in major metropolitan centers through programs like SCOOT (Split Cycle Offset Optimisation Technique), a system that responds to fluctuations in traffic flow through detectors embedded in the road. It helps the FAA’s National Command Center in Herndon, Va., predict how a flight delay in Seattle will change departure times in New York. Many brokerage companies use AI programs to better time stock trades. Your local utility company uses AI to operate more safely and efficiently, and since the launch of the Apple iPhone 4S in October, millions of people around the world are using an AI program called Siri to find nearby restaurants, schedule meetings and more.
“In economic terms, automation in general should be seen as a leveraging factor that amplifies the output of workers,” says Rod Furlan, an AI researcher and machine-learning expert based in Vancouver. “Thanks to the availability of legal software, one lawyer can do today work that required a team of assistants 10 years ago. Ten years from now, an individual lawyer may be able to service as many cases as a small firm does today, all thanks to AI advancements. Going forward, we can expect to do less boring work and have more time for truly intellectual tasks which are less likely to be automated in the near term.”
Artificial intelligence will make its way into many more areas of life and become much more robust, experts predict. This phenomenon is driven by several factors. The first is the acceleration of computational capability. The amount of data you can squeeze onto an integrated circuit continues to double every 18 months or so, and that will continue until around 2015, at which point the doubling effect will slow, but still continue.
The second factor driving greater investment in – and wider adoption of – AI is simple information overload. We’re creating far too much data to process effectively without robust computational tools. Every day, the U.S. Air Force processes 1,500 hours of full-motion video and 1,500 still images taken from aerial drones operating in Afghanistan and Iraq. Artificial intelligence helps analysts extract meaningful insight from that information to better execute operations in real time. Over the next two decades, that technology will get better, but it’s already leaving the confines of military use and finding new markets in the civilian world.
In 2009, the Fraunhofer Institute for Intelligent Analysis in Germany created a search engine that combs through broadcasts for spoken words, so if you want to know exactly what one candidate said during a particular debate, you won’t have to sift through news articles. Instead, you can search the debate the same way you search for information on Google today. As more people spend more time in the presence of microphones, and as more footage from those devices goes online, that will allow someone with a smart phone to look up any recorded conversation between two people that occurred anywhere a microphone was present.
“Information is already being aggregated by law-enforcement organizations across social networking sites, surveillance cameras, electronic transactions and online data for purposes of national security and law enforcement,” says Dylan Glas, an MIT-trained artificial intelligence researcher working in Japan. As computing processing power and software improves, data mining with the help of AI programs will become “an essential tool in business as well, enabling predictive modeling of consumer behavior and highly targeted marketing to individuals.”
One application of AI is making predictions, based on large data sets, about how people will behave and how to give people more of what they truly want. Google CEO Eric Schmidt has said that the perfect search engine anticipates what you’ll be searching for before you go to your keyboard. Apply that principle to the entire world. Imagine cars, stores, scales and other objects you encounter anticipating your needs, in concert with the needs of others, and trying to give you what it thinks you need before you ask. All this happens invisibly as we merge the massive amount of data we can now collect with emerging technological tools.
We can use that new real-time understanding to be more efficient with resources. This is the central idea behind the smart-grid proposal endorsed by the Obama administration. A smart grid would predict individual future energy demands based on data received from sensors across millions of homes, eliminating waste and bringing down energy costs for everyone.
In the next decade, AI could accelerate the hunt for new medicines and cures. With rapidly aging populations in the Untied States and Europe, there is, perhaps, no need that is more pressing. Some economists anticipate that spending on elder care in the United States will double by the middle of the century. Artificial intelligence might help researchers make better use of the massive amounts of medical and biophysical data to quicken the pace of discovery.
“If you can sequence everyone’s genome and do machine learning on the data, making an existing algorithm more useful, you might, say, find a gene correlated across the board for obesity, hypertension and cancer,” says Eliezer S. Yudkowsky, an AI theorist affiliated with the nonprofit Singularity Institute. “If we were a sane society, we could use that capability to output lots of drugs to make people healthier, happier and smarter.”
One of the most noticeable applications of artificial intelligence in the near future will be autonomous vehicles. Progress in this area has been quietly building for decades, but it reached an important milestone in 2005 with the debut of Stanley, a self-driving car (in photo below) that successfully traveled 132 miles in six hours and 54 minutes, winning the inaugural Defense Advanced Research Projects Agency (DARPA) Grand Challenge event. Stanley’s race time represents a four-orders-of-magnitude improvement over the self-driving cars that Stanford researchers were working with in 1979, which required six hours to travel 20 meters. Researchers expect autonomous driving robots to develop much faster in the next 30 years.
Private companies are making large investments in artificially intelligent vehicles. Last May, Google began to aggressively lobby the state of Nevada to allow self-driving cars on the state’s roads. At the core of the current Google proposal is an elaborate radio-frequency identification system that would enable robotic vehicles to sense the presence of other cars and coordinate movements. There’s nothing in the Google proposal that doesn’t already exist, from a technical standpoint. Cars already communicate with tollbooths via radio, but not yet with other cars.
It may be a long time before such car-to-car collaboration eliminates the need for traffic lights and speed limits. In the meantime, vehicles that could better negotiate position, speed and distance with one another would improve highway safety, whether or not those cars have human operators. In the future, such social software may allow groups of robots to organize themselves into operating structures most appropriate for a given task. Under certain conditions, the robots may choose a single leader. In other circumstances, a highly democratic structure may emerge. Robots will be able to select and implement whatever organizational method suits the situation. This is something humans take forever to do but that robots could do fairly quickly.
“Although the automobile industry is a notoriously slow adopter of new technologies, we are already seeing the implementation of pre-crash and collision-avoidance systems in recent cars,” Glas says. “Now that sensor data is available and software is able to control the actuation of steering, acceleration and brakes, there is no longer any major technical barrier .... The AI under the hood will certainly advance to the point where the driver takes on more of a supervisory role. In terms of traffic safety in everyday life, this will have an enormous impact.”
In the workplace, AI may boost worker productivity to a degree not seen since the invention of the personal computer. The economic impact is less easy to predict. Historically, large breakthroughs in productivity, roughly defined as the amount of output per worker per hour, are followed by increases in employment. Companies lay off workers and invest in tools and innovation to make more goods with less money. This increases profit, furthering demand for goods across the economy and necessitating the need to hire more workers. This cycle, in part, is why more Americans had jobs at the end of each decade of the 20th century than they did at the beginning of that decade. Historically, new technology creates more jobs than it destroys.
Even so, AI researchers and management experts disagree about the influence AI will have for broader employment and productivity.
“AI can certainly lead to quality-of-life improvements, but how does that actually translate into more jobs?” asks Yudokwsky, who says he doesn’t believe increased worker productivity alone will put the U.S. economy back on a path of high growth. “Advances in AI technology tend to substitute for unskilled human labor. Unless the rest of the economy pulls its act together, AI won’t pull it out of its funk.”
Furlan says that as more businesses embrace aggressive automation opportunities through AI and advanced robotics, we’re likely to see more companies that, like Google, have an astronomical revenue-per-employee ratio. He adds that he’s still “bullish” on AI and is confident that businesses and individuals will be able to adapt to the new era of increased worker capability.
In their most recent book, Race Against the Machine: How the Digital Revolution is Accelerating Innovation, Driving Productivity, and Irreversibly Transforming Employment and the Economy, MIT scholars Andrew McAfee and Erik Brynjolfsson write, “There is no law that says everyone, or even most people, automatically benefit from technological progress.” They argue that digital technologies advance much faster than organizations. While artificial intelligence and other productivity boosters create tremendous value, that value won’t necessarily trickle down to workers as it did during the 20th century. “As technology accelerates,” they say, “so will the economic mismatches, undermining our social contract and ultimately hurting both rich and poor.”
Roboticist Rodney Brooks is more optimistic. Brooks is most famous for the creation of the Roomba robot vacuum cleaner, but he’s also the creator of the PackBot, a bomb-disposal robot that’s been in use in Iraq and Afghanistan since the start of U.S. operations there.
In 2008, Brooks left his position at MIT and founded Heartland Robotics, a small start-up with financial backing from Amazon founder Jeff Bezos and Silicon Valley venture capital firm Charles River Ventures. Heartland seeks to develop an entirely new type of versatile factory robot.
“Our robots will be intuitive to use, intelligent and highly flexible,” Brooks says. “They’ll be easy to buy, train and deploy, and will be unbelievably inexpensive. Heartland Robotics will change the definition of how and where robots can be used, dramatically expanding the robot marketplace.”
At a recent National Academy of Engineering event in Washington, Brooks expressed confidence that innovation in AI, and particularly in robotics, can help the United States resurrect its reputation as a manufacturing powerhouse. “I think we need a new vision for American manufacturing,” he said. “We need skilled workers who are producing both high-value and mass-market products. The way I see us doing that is through robots, allowing them to take over the simple parts of the simple tasks. Let the robots do the dumb stuff.”
Rodney Brooks with his Obrero machine (photo at left), the basis for the soon-to-be-released Heartland factory robot Brooks and other parties associated with Heartland aren’t speaking to the media about the company’s current work. “We’re in stealth mode,” he wrote in an email. But published reports in IndustryWeek, and previous statements by Brooks, suggest that the Heartland project could forever change the way robots are used in manufacturing.
Today’s factory robots require a large capital investment up front, can only perform specific and limited tasks, and require extensive integration. So when you introduce a new industrial robot into a factory setting, you have to remake large areas of the assembly line around what the robot can do and the way it operates. This, in part, is why businesses that use industrial robots make products that are high-value and don’t change much over time, like cars.
To make products that change very quickly from season to season, like toys or even electronics, you need a human work force, preferably a very inexpensive one, because the profit margin on these items is usually low. That’s why so much manufacturing migrates to places where human labor is cheap and plentiful, despite 30 years of automation advances in more developed economies.
The Heartland robot, according to reports, would be smaller and more versatile than previous factory robots. It would consist of an arm with limited haptic (sensing or tactile) capability, a camera, and AI enabling the machine to complete a wide variety of tasks, from inspecting products for flaws to packaging them for shipment. Most importantly, Heartland has reportedly established a retail price target of $5,000.
Andrew McAfee says the Heartland robot would make it possible for small-business owners to “quickly set up their own highly automated factory, dramatically reducing the costs and increasing the flexibility of manufacturing.”
“There are 300,000 companies with less than 500 employees involved in manufacturing” in the United States, Brooks says. “If you go out there and look at them, walk around their floors, a lot of them are operating just the same way that they were 50 years ago.”
Brooks’ vision of the future is one where nimble garage enterprises in the Midwest and elsewhere compete on a global scale against factories in Beijing. In this scenario, government and large institutions still have a vital role to play, funding basic research and development and investing in a work force that can take advantage of ever more complex technologies. But this great leap forward will be accelerated not by public money or institutional muscle so much as by the proliferation and better application of intelligence, computational and otherwise.
The third Industrial Revolution will be entrepreneurial in character, endowing individuals with greater capability and demanding more from them in return, in terms of the time and effort it takes to acquire technological skills. If we, as a nation of individuals, can will ourselves to embrace the tools at our disposal as well as the challenges and opportunities those tools present, the future can be incredibly bright.