Genially sans titre

1956-1975

1981-1985

1995-2010

1976-1980

1985-1995

2010-

-1950

The Evolution of Artificial Intelligence Throughout History

Instructions: Explore the timeline by clicking on the different images and reveal details about the crucial milestones in the history of AI!

Start

1956-1975

1981-1985

1995-2010

1976-1980

1985-1995

2010-

-1950

The Evolution of Artificial Intelligence Throughout History

1956-1975

1981-1985

1995-2010

1976-1980

1985-1995

2010-

-1950

The Evolution of Artificial Intelligence Throughout History

Pre-AI Era:Artifical Intelligence is deeply linked to computers, so it is not a surprise that AI was not a subject before the 50s as the computers appeared only during the late 40s. The idea of automation is however already present, mainly through analog devices using feedbacks from sensors as described by Wiened under the name cybernetics, for instance to control heaters or rockets. At the same time, the idea of extracting knowledge from data is developing through the Data Analysis framework of Tukey or the classical statistic framework evolved from Fisher with numerous applications in health science.

1956-1975

1981-1985

1995-2010

1976-1980

1985-1995

2010-

-1950

The Evolution of Artificial Intelligence Throughout History

AI Era:In 1950, Alan Turing already asks a fundamental question “Can computer think?”, or rather, can they imitate any human behavior?  This can be seen as the birth of Artificial Intelligence as a topic, even if the name AI itself is only coined in 1946 by McCarthy, Minsky and Shannon.  More precisely, Artificial Intelligence was defined as “the science and engineering of making intelligent machines”. It was restricted mostly to the idea of what is called “symbolic AI” in which the rules of the world are given by experts and used by computers to act.  Those principles have been applied to games (checkers for instance) or to check simple mathematical proofs. In parallel, the first artificial neuron, which uses computation and feedbacks from experiments to learn how to perform a task, is developed. Due to scarce computing resources, this direction did not obtain good results and was seen as pointless by the symbolists.

The Evolution of Artificial Intelligence Throughout History

1956-1975

1981-1985

1995-2010

First AI Winter: After a period of growth and progress, the symbolic AI period had reached a peak: problems that have been solved are seen as simple and almost no progress are made on the one seen as the most promising, like computer vision for instance This is a period of disillusion, called the (first) AI Winter, in which there is much less funding, and thus research, on AI.

1976-1980

1985-1995

2010-

-1950

1956-1975

1981-1985

1995-2010

1976-1980

1985-1995

2010-

-1950

The Evolution of Artificial Intelligence Throughout History

Expert System Era:During the 70s, database systems, in which a lot of information can be stored, were developed. The stored data was used mainly to track transactions and compute their effects. AI makes a comeback at that time, as symbolic AI can be used to applied complex rules, given by experts, to the same data.  As powerful as this idea can be, it was limited by the computation power available at that time and the difficulty to define the rules. 

1956-1975

1981-1985

1995-2010

1976-1980

1985-1995

2010-

-1950

The Evolution of Artificial Intelligence Throughout History

Second AI Winter: Similarly to the first AI Winter, AI entered a period of disillusionment and reduced funding without any noticeable breakthroughs. Note that there were advancements made during this time, most of them we are using today such as machine learning and neural networks. However, these advancements remained largely unnoticed in the background.

1956-1975

1981-1985

1995-2010

1976-1980

1985-1995

2010-

-1950

The Evolution of Artificial Intelligence Throughout History

Machine Learning Era:We had to wait for the computational progress of the 90s to witness a significant shift in AI from rule-based approaches to learning-based ones. This shift, combined with the availability of larger and larger datasets, led to the first successes of AI in areas such as computer vision and chess playing.During this period, clever learning techniques matured, and there was a focus on adapting these techniques to specific use cases. Most successes were observed in domains where information could be easily summarized in a structured format, excluding text and images, for instance.While symbolic AI continued to exist, it became less prominent and visible during this time.

1956-1975

1981-1985

1995-2010

1976-1980

1985-1995

2010-

-1950

The Evolution of Artificial Intelligence Throughout History

Deep Learning Era:From 2010, an old idea, neural networks, (re)invented in the 80s and extended in the 90s becomes the state-of-the-art approach in most applications involving text and images, known as deep learning. This shift was mainly due to significant advancements in computation and data scalability, facilitated by the development of the graphical card. As you can see every day, the results are impressive: automatic translations are much better than what they used to be, we can generate images from a description or interact with a chatbot… However, deep learning is not the only method used. Classical machine learning algorithms are still used for data stored in table. Even for text and images, symbolic AI continues to play a role in the search for methods to better control outputs and incorporate prior knowledge, contrasting the principles of learned rules.