Over the past several decades, artificial intelligence has been besieged by rounds of hype that over-promised, under-delivered and nearly killed the field. Once more, Gartner positioned machine learning and deep learning at the top of its hype cycle, and all the publicity and potential often leads to a “trough of disillusionment” if the technology fails to deliver. Now that AI is reaching a tipping point of market acceptance, it’s important to be cautious and not repeat past mistakes.
In “Artificial Intelligence – The Revolution Hasn’t Happened Yet,” University of California at Berkeley professor Michael I. Jordan injects such a note of caution. “The idea that our era is somehow seeing the emergence of an intelligence in silicon that rivals our own entertains all of us —enthralling us and frightening us in equal measure,” he writes. “Whether or not we come to understand intelligence any time soon, we do have a major challenge on our hands in bringing together computers and humans in ways that enhance human life.”
Tools have played a critical role in the evolution of humans since our ancestors first developed stone tools a few million years ago. “We shape our tools and they in turn shape us,” observed noted author and educator Marshall McLuhan in the 1960s.
Similarly, the machines of the 21st century digital economy are making up for our cognitive limitations, augmenting our intelligence, problem solving capabilities and ability to process vast amounts of information. Machine learning and deep learning are the latest examples of tools that are helping us cope with and take advantage of the huge amounts of information all around us.
Mr. Jordan argues that we are witnessing the creation of a new branch of engineering that will help us in the development of AI systems and applications, much as civil and mechanical engineering helped us develop tall buildings and airplanes over a century ago. Before the advent of these engineering disciplines, buildings and bridges were developed in fairly ad-hoc ways, and were much less safe and subject to collapsing in unforeseen ways. Over time, engineering advances have led to foundational scientific principles, development practices and building blocks that significantly increased their safety and productivity.
AI is in this ad-hoc, early stage. Tools and best practices have started to emerge, as well as a number of sophisticated mathematical techniques such as those underly deep learning. However, “what we’re missing is an engineering discipline with its principles of analysis and design…”, Mr. Jordan notes. “Moreover, since much of the focus of the new discipline will be on data from and about humans, its development will require perspectives from the social sciences and humanities.”
In the 1950s, the founders of AI aspired to build machines with human-level intelligence, believing such “human-imitative AI” would be achieved within a generation. Sixty years later, that high-level reasoning and thought has yet to materialize, Mr. Jordan says. “The developments which are now being called AI arose mostly in the engineering fields associated with low-level pattern recognition and movement control, and in the field of statistics —the discipline focused on finding patterns in data and on making well-founded predictions, tests of hypotheses and decisions.”
Indeed, much of the progress has come in the form of intelligence automation, in which computers and data are used to create IT-based tools that augment human capabilities. In addition, the emergence of the Internet of Things and applications like smart cities and smart manufacturing, are giving rise to system-wide intelligent infrastructures.
For the foreseeable future we have to rely on intelligence augmentation, intelligent infrastructures and similar engineering approaches to continue AI’s advances. “We are very far from realizing human-imitative AI aspirations,” Mr. Jordan writes. “Unfortunately the thrill (and fear) of making even limited progress on human-imitative AI gives rise to levels of over-exuberance and media attention that is not present in other areas of engineering.”
Moreover, the most challenging problems areas—healthcare, transportation, finance, education, government—require highly complex engineering and systems-oriented advances. When addressing those challenges, it’s possible a focus on human-imitative AI might be a distraction, Mr. Jordan says.
“While the humanities and the sciences are essential as we go forward, we should also not pretend that we are talking about something other than an engineering effort of unprecedented scale and scope — society is aiming to build new kinds of artifacts. These artifacts should be built to work as claimed.”
While industry may drive many AI developments, academia will play an essential role by providing new technical ideas and bringing researchers together from across disciplines, he says. That includes the much-needed perspective of experts in the social sciences, cognitive sciences and humanities, as well as those in computational and statistical disciplines.
In the end, concludes Jordan, a new branch of engineering is being created that blends engineering with data-focused and learning-focused disciplines, as well as humanities and social sciences. “We have a real opportunity to conceive of something historically new — a human-centric engineering discipline.”