AI is everywhere! But where does it come from? This blog provides clarity by telling the exciting story of artificial intelligence.

The Birth of AI

John McCarthy, one of the founders of Artificial Intelligence.

Artificial Intelligence, or AI, is a broad field of computer science aimed at creating systems capable of performing tasks that typically require human intelligence. The term AI, was first mentioned by John McCarthy in 1955, as "the science and engineering of making intelligent machines." Convinced that artificial intelligence has a bright future, he organizes a two-month brainstorming session with ten other scientists. His proposal letter reads "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." The Dartmouth Workshop finally took place in 1956 and is considered the birth of AI as a field of research.

Early Hopes and the AI Winter

The hopes of matching human intelligence with artificial intelligence in just a few years are high, and there is no shortage of creative ideas. In 1957, Frank Rosenblatt introduced the first simulation of the human brain, the multilayer perceptron. This approach belongs to the field of machine learning, a subset of AI that enables computers to learn from data. What Rosenblatt couldn't foresee was that his approach would eventually lead to a breakthrough in AI research. However, at the time, there was a lack of mathematical methods and sufficient computing power to fully realize its potential.

Limited to static algorithms, early applications such as Knowledge-Based Systems emerged. These systems combined facts stored in a knowledge base with if-then rules to create expert systems that, in theory, could replace human experts. However, the systems were rigid, and adjustments to their logic had to be made manually. The anticipated successes never materialized, and the term "artificial intelligence" faded into obscurity during the 1970s – leading to what is now referred to as the AI winter.

The Breakthrough of Machine Learning

While AI was stagnating, progress was being made in other areas. In the 1980s, the discovery of backpropagation, a mathematical method for optimizing neural networks, gave machine learning a second chance. However, processing the large datasets needed for these systems proved challenging due to the limited computing power of the time.

By the late 1980s, the field began to pick up momentum again. In 1989, a machine learning algorithm successfully classified handwritten digits. In 1995, a semi-autonomous car drove across the United States, and in 1996, IBM's Deep Blue defeated reigning world chess champion Garry Kasparov.

The Era of Deep Learning

However, the limitations of traditional machine learning methods soon became evident, and new approaches were needed. This is where deep learning came into play – a subset of machine learning that, through the development of deep neural networks, ushered in a new era of AI research. These networks made it possible to solve far more complex problems than classical methods ever could.

A protein predicted by Alpha fold

CC BY 4.0

In 2012, a deep Learning architecture designed to analyze images called convolutional neural network (CNN) won the ImageNet competition, revolutionizing image recognition. In 2014, the new GAN architecture turned the tables, and instead of just extracting information from images, it managed to create images from information. In 2018, DeepMind's AlphaFold made a major breakthrough in natural sciences by winning the CASP competition, solving a long-standing challenge in biology. By accurately predicting protein structures, AlphaFold revolutionized our understanding of biological processes, with far-reaching implications for drug discovery and disease research. These breakthroughs solidified deep learning's role in AI innovation.

The path to human like intelligence

Most of the AI invented in the last decades operates as Artificial Narrow Intelligence (ANI), specialized on a single task like playing chess. However, recent advances in deep learning are opening up new opportunities to tackle more general problems. In other words, we are looking for Artificial General Intelligence (AGI). Large language models (LLMs), such as OpenAI's GPT-3 (2020) are a good example. These models are trained on vast datasets, enabling them to perform a wide variety of tasks, ranging from answering questions to writing code. Unlike ANI, AGI aspires to mimic human intelligence by learning, reasoning, and adapting across different domains. This trend brings AI closer to human-like flexibility, though true AGI remains a distant goal.

The future of AI

Artificial intelligence has come a long way since 1956. And while AI was already outperforming humans in certain tasks in the 1990s, the hype has peaked today because interacting with AI feels so natural and results are almost too realistic. This not only generates excitement and dreams of the future, but also fears. Voices are raised that AI is taking big steps towards artificial superintelligence and its domination of the world is getting closer. At the same time, the economic value it generates often doesn't match the excitement. Only time will tell how AI will affect humanity. But one thing is certain: we are on the verge of a new era.