History of AI: Part Two — The Leap to practical applications (1960s)

The 1960s saw the field of Artificial Intelligence shift from theoretical discussions to producing actual, measurable results.

This period emerged as an important phase in the development of AI. It was characterized by scientific progress and a heightened sense of anticipation among researchers. It was an era where AI started to weave into the fabric of daily living, progressively becoming a fundamental element of human experience. Computers, once mere tools for computation, were now solving algebraic problems, proving geometric theorems, and even venturing into the domain of understanding and processing human language.

This era saw the blossoming of various approaches in AI research. One of the most notable was the paradigm known as reasoning as search. AI programs adopted this approach to methodically work through problems, akin to navigating a maze, searching for solutions, and retreating when faced with dead ends. This method, however, came with its challenges, particularly the combinatorial explosion: the vast number of potential paths to a solution. To combat this, researchers employed heuristics, or rules of thumb, to streamline the process.

The foundation of Neural Networks and NLP

Following the groundwork laid in the 1940s by McCulloch and Pitts, the 1960s saw important developments in neural networks, with advances in hardware capable of simulating these networks. Innovators like Frank Rosenblatt and Bernard Widrow were forefront in this area, creating machines such as the Perceptron and ADALINE with adjustable weights, crucial for neural networks. These efforts marked a transition from theoretical exploration to practical, physical models.

Natural Language Processing (NLP) also made substantial strides. Early achievements like Daniel Bobrow’s STUDENT program, which solved algebraic problems, and Joseph Weizenbaum’s ELIZA, capable of holding conversations, demonstrated AI’s improved proficiency in human language interaction. ELIZA was particularly groundbreaking, with conversations so realistic that some users believed they were speaking to a human.

The late ’60s saw AI research pivot with the concept of micro-worlds, led by Marvin Minsky and Seymour Papert at MIT. This method involved focusing on simplified environments to understand complex concepts, leading to advancements in machine vision and robotics, as seen in Terry Winograd’s SHRDLU program, capable of understanding and executing commands in a simulated environment.

Globally, AI’s influence spread to Japan with the WABOT project, which resulted in WABOT-1, the first full-scale intelligent humanoid robot. Its abilities in movement, object handling, and basic communication were groundbreaking in robotics.

Funding was crucial for AI’s progress in the 1960s. Significant investment from agencies like DARPA supported numerous AI projects and research initiatives. This financial backing enabled institutions like MIT, CMU, Stanford AI Project, and Edinburgh University to become AI research leaders. DARPA’s funding strategy focuses on individuals and ideas rather than specific projects. This, in turn, fostered a fertile environment for innovation and experimentation.

The 1960s were an encouraging period for AI, moving beyond technological advances to redefine machine capabilities. This era marked the transition of AI from theoretical concepts to practical applications, setting the stage for the technologies we depend on today.