From “ARTIFICIAL INTELLIGENCE AND LIFE IN 2030”.
Artificial Intelligence (AI) is a science and a set of computational technologies that are inspired by—but typically operate quite differently from—the ways people use their nervous systems and bodies to sense, learn, reason, and take action.
While impressive, these technologies are highly tailored to particular tasks. Each application typically requires years of specialized research and careful, unique construction.
People’s future relationships with machines will become ever more nuanced, fluid, and personalized as AI systems learn to adapt to individual personalities and goals.
As a society, we are now at a crucial juncture in determining how to deploy AI-based technologies in ways that promote, not hinder, democratic values such as freedom, equality, and transparency.
Well-deployed AI prediction tools have the potential to provide new kinds of transparency about data and inferences, and may be applied to detect, remove, or reduce human bias, rather than reinforcing it.
In many realms, AI will likely replace tasks rather than jobs in the near term, and will also create new kinds of jobs. But the new jobs that will emerge are harder to imagine in advance than the existing jobs that will likely be lost. AI will also lower the cost of many goods and services, e ectively making everyone better o . Longer term, AI may be thought of as a radically di erent mechanism for wealth creation in which everyone should be entitled to a portion of the world’s AI-produced treasures. It is not too soon for social debate on how the economic fruits of AI technologies should be shared.
Machine learning has been propelled dramatically forward by impressive empirical successes of arti cial neural networks, which can now be trained with huge data sets and large-scale computing.
Reinforcement learning is a framework that shifts the focus of machine learning from pattern recognition to experience-driven sequential decision-making. It promises to carry AI applications forward toward taking actions in the real world. While largely con ned to academia over the past several decades, it is now seeing some practical, real-world successes.
Public policies should help ease society’s adaptation to AI applications, extend their bene ts, and mitigate their inevitable errors and failures.
A major focus of current efforts is to scale existing algorithms to work with extremely large data sets. For example, whereas traditional methods could afford to make several passes over the data set, modern ones are designed to make only a single pass; in some cases, only sublinear methods (those that only look at a fraction of the data) can be admitted.
Whereas traditional machine learning has mostly focused on pattern mining, reinforcement learning shifts the focus to decision making, and is a technology that will help AI to advance more deeply into the realm of learning about and executing actions in the real world. It has existed for several decades as a framework for experience-driven sequential decision-making, but the methods have not found great success in practice, mainly owing to issues of representation and scaling. However, the advent of deep learning has provided reinforcement learning with a “shot in the arm.”
The deep learning revolution is only beginning to influence robotics, in large part because it is far more difficult to acquire the large labeled datasets that have driven other learning-based areas of AI.
Reinforcement learning, which obviates the requirement of labeled data, may help bridge this gap but requires systems to be able to safely explore a policy space without committing errors that harm the system itself or others.
Overall trends and the future of AI research: The resounding success of the data-driven paradigm has displaced the traditional paradigms of AI. Procedures such as theorem proving and logic-based knowledge representation and reasoning are receiving reduced attention, in part because of the ongoing challenge of connecting with real-world groundings. Planning, which was a mainstay of AI research in the seventies and eighties, has also received less attention of late due in part to its strong reliance on modeling assumptions that are hard to satisfy in realistic applications. Model-based approaches—such as physics-based approaches to vision and traditional control and mapping in robotics—have by and large given way to data-driven approaches that close the loop with sensing the results of actions in the task at hand. Bayesian reasoning and graphical models, which were very popular even quite recently, also appear to be going out of favor, having been drowned by the deluge of data and the remarkable success of deep learning.
Over the next fteen years, the Study Panel expects an increasing focus on developing systems that are human-aware, meaning that they speci cally model, and are speci cally designed for, the characteristics of the people with whom they are meant to interact.
We encourage young researchers not to reinvent the wheel, but rather to maintain an awareness of the signi cant progress in many areas of AI during the rst fty years of the eld, and in related elds such as control theory, cognitive science, and psychology.
Autonomous transportation will soon be commonplace. As cars will become better drivers than people, city-dwellers will own fewer cars, live further from work, and spend time differently, leading to an entirely new urban organization. Current cars can park themselves, perform adaptive cruise control on highways, steer themselves during stop-and-go traffic, and alert drivers about objects in blind spots during lane changes. Vision and radar technology were leveraged to develop pre-collision systems that let cars autonomously brake when risk of a collision is detected. Deep learning also has been applied to improve automobiles’ capacity to detect objects in the environment and recognize sound.
Google’s self-driving cars, which have logged more than 1,500,000 miles (300,000 miles without an accident),33 are completely autonomous—no human input needed. Tesla has widely released self-driving capability to existing cars with a software update.34 Their cars are semi-autonomous, with human drivers expected to stay engaged and take over if they detect a potential problem. It is not yet clear whether this semi-autonomous approach is sustainable, since as people become more con dent in the car’s’ capabilities, they are likely to pay less attention to the road, and become less reliable when they are most needed.
Advances in perception will be followed by algorithmic improvements in higher level reasoning capabilities such as planning. A recent report predicts self-driving cars to be widely adopted by 2020. And the adoption of self-driving capabilities won’t be limited to personal transportation.
Self-driving cars will eliminate one of the biggest causes of accidental death and injury in United States, and lengthen people’s life expectancy. On average, a commuter in US spends twenty- ve minutes driving each way.
The increased comfort and decreased cognitive load with self-driving cars and shared transportation may a ect where people choose to live.
Shared autonomous vehicles—people using cars as a service rather than owning their own—may reduce total miles, especially if combined with well-constructed incentives, such as tolls or discounts, to spread out travel demand, share trips, and reduce congestion.