A new MIT artificial intelligence project, the Mind Machine Project (MMP), recently began research that will “reconcile natural intelligence with machine intelligence, and in doing so develop and engineer a class of intelligent machines.” The researchers believe that many aspects of the foundations of AI have been rooted in mistakes made over the past generation of AI research. Therefore, they think that combining “current advances in each of these areas with insights from their roots, it will be possible to fulfill the early vision that lies at their intersection.” As part of the process they will reexamine assumptions about AI and attempt to unveil new answers to fundamental questions that fall within all aspects of AI study.
What exactly is artificial intelligence? As Stanford’s John McCarthy provides answers to basic questions on AI, he states that AI is “the science and engineering of making intelligent machines, especially intelligent computer programs. It is related to the similar task of using computers to understand human intelligence, but AI does not have to confine itself to methods that are biologically observable.” People regard the 1956 Dartmouth summer conference as the birth of AI as a concept and field of study. The original AI project proposal contained lofty goals and bold aspirations:
The study is to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves. We think that a significant advance can be made in one or more of these problems if a carefully selected group of scientists work on it together for a summer.
Regardless of how one interprets a “significant advance,” it’s mind-boggling to imagine that the original AI research team believed that substantial progress in the field could be made in a two month study. Continued optimism in the early years of AI research was eventually hindered by fundamental problems in the field of AI that restricted possible advances, such as limited computer power, interactibility and the combinatorial explosion problem, commonsense knowledge and reasoning, the frame problem, Moravec’s Paradox, and the qualification problem.
In the first half of the 1980s, the implementation of expert systems by major corporations lead to a new area of focus for AI researchers. These computer programs answer questions and solve problems using logic rules that stem from the knowledge of experts. Unfortunately, expert systems are limited to a specific domain of knowledge, so they avoid attempts to tackle the commonsense knowledge problem. The key significance of expert systems in the history of AI, is that they were the first practical and beneficial products in this field.
Lamentably, the AI success stories in the early 1980s were followed by an AI winter, a period of reduced funding in the field. In essence, this AI winter was akin to the recent real estate collapse or the dot com bubble of the earlier part of this decade. A period of heighten enthusiasm and optimism in the field of AI generated ambitious goals that, in hindsight, were clearly unrealistic. Eventually, people outside the AI community that provided funding, such as government agencies and corporations, lost their patience as certain expectations could not be met and significantly reduced investments into AI research. Thus, busting the AI boom and bringing the onset of the AI winter. Although it seems that this still continues to today, in recent years it has been believed that the spring for AI is near.
In this context, the new MIT MMP, appears especially promising and may result in a season of change for the area of AI. The stated goals for the MMP focus in three areas: Mind (models for thought), Memory (accumulating and using experience), Body (scalable substrates to embody intelligence), and Brain/Intent (looking for advanced applications of these technologies, such as “non-chemical based” solutions for psychiatric treatments and brain prostheses).
In the 5000 year timeline of AI and related study, the areas of mind, memory and body have been studied, defined, debated, understood, reworked, reviewed, rewritten and redefined. The current MMP research team will rethink past approaches to AI. Part of this process includes taking a big picture approach to develop sound theory and create useful products and programs. Too often in the past, AI projects focused on quickly getting software products to the market without considering the ties to possible theoretical advances. Basically, the new team wants a holistic approach to tackling the challenges of AI. They plan on making great gains in the field through the implementation of a comprehensive research strategy. Maybe this will be the start of a new age of AI. At a recent reunion of veritable legends of AI researchers, the pioneers expressed optimism for the future of AI.
Is the development of the true AI soon? Could the MMP or other researchers create an AI that could pass the ultimate Turig Test, a test that distinguishes AI from a human through the course of conversation? Will we soon see AI that can interact with us in a way only seen in science fiction movies? Is a scene, like the one below in our near future?