A Brief History of Artificial Intelligence: AI Development Over Time

The field of artificial intelligence (AI) has fascinated scientists, computer scientists, and digital technology enthusiasts for decades. Its emergence and evolution are a tale of technological breakthroughs, surprising discoveries, and moments of stagnation.

In this article, we'll examine the development of AI, from its origins to its contemporary applications, highlighting the key events, people, and technologies that have shaped the field.

What is artificial intelligence?

Today, artificial intelligence (AI) refers to the ability of machines and computer systems to simulate human intelligence and the workings of the human brain. AI enables robots and machines to:

• problem-solving,

• decision-making,

• natural language processing,

• independent programming,

• predicting trends,

• generating content and graphics,

• continuous learning.

A brief history of artificial intelligence - milestones from the Turing Test to the present day.

Before we get to the intriguing stories, let's get to know the most important dates and events:

• 1950: Alan Turing proposes the so-called "Turing Test" as a way to assess the ability of machines to imitate human thought. This test also laid the foundation for the later creation of personal computers.

• 1956: Scientific conference in Darmstadt, where the term "artificial intelligence " was first used. John McCarthy, an American scientist, coined it and has since been called the father of artificial intelligence.

• 1966: Eliza, a precursor to today's chatbots, was created by Joseph Weizenbaum.

• 1980: A shift towards machine learning occurs. The first so-called expert systems were created.

• 2000: The rebirth of AI. After the difficult end of the 20th century, the development of artificial intelligence systems began to accelerate rapidly.

• 2020: Advanced language models like ChatGPT are breaking new ground and starting a revolution in content creation.

• 2023: Artificial intelligence becomes a key tool in fields such as healthcare, robotics, and autonomous vehicles.

The birth of artificial intelligence: Alan Turing and his vision.

The history of AI begins with the work of British mathematician, computer scientist, and military officer Alan Turing, who proposed a groundbreaking test in 1950.

Its purpose was to assess whether a machine could think like a human. In a so-called blind test, a judge conducts a dialogue with both the machine and the human, unaware of which interlocutor is which. If the judge is unable to distinguish between the interlocutors, the machine is deemed to have passed the intelligence test.

Why is this test so important?

• This was the first step in the practical definition of machine intelligence.

• Its importance lies in setting standards for AI research.

• Inspired the further development of IT-related technologies.

Turing's research laid the foundations for modern information technology.

Dartmouth Conference – the beginning of AI research.

In 1956, during an 8-week summer research project at Dartmouth in the United States, John McCarthy introduced the term "artificial intelligence".

This event is considered the birth of AI as a distinct field of science. A group of scientists, including McCarthy, Marvin Minsky, Nathaniel Rochester, Claude Shannon, Allen Newell, and Herbert A. Simon, decided to begin research into creating machines capable of intelligent behavior. Thus began the development of artificial intelligence.

At this point, it's worth devoting a few more words to Professor McCarthy, who became a key figure in the development of computer science, and whose name is included on the official list of pioneers in the field. He created, among other things, the LISP programming language and co-created ALGOL. He also played a key role in the invention of time-sharing systems.

His colleague from the university, Lester Earnest, commented on this in an interview:

The internet wouldn't have emerged so quickly if it weren't for John spearheading the development of time-sharing systems. We're constantly coming up with new names for time-sharing . They started calling it servers... Now we call it cloud computing. It's still just time-sharing. John started it.

As for other pioneers, besides J. McCarthy, it is also worth mentioning Arthur Samuel, who in 1959 was the first to create a program that played chess.

When the first systems and algorithms for machines and computers began to be programmed.

Research into artificial intelligence began in the 1950s. In the 1950s, 1960s, and 1970s, artificial intelligence algorithms were developed that allowed machines to learn and solve simple logical problems.

Logic Theorist.

Logic Theorist, created in 1955 by Allen Newell, Clifford Shaw, and Herbert Simon, is often called the first AI program. It was a computer program designed to imitate human reasoning by proving mathematical theorems in first-order logic.

The Logic Theorist proved theorems from Bertrand Russell and Alfred North Whitehead's Principia Mathematica, which was revolutionary for the time. Instead of relying on a structured set of rules, he used heuristics, or general strategies for approaching problems. He not only verified whether a given proof was correct, but could also generate new proofs for a given theorem.

Logic Theorist marked a significant milestone in the evolution of artificial intelligence, which has taken many unexpected turns. It confirmed that machines could imitate human thought processes in a field as complex as mathematics.

General Problem Solver (GPS).

Moving on, it's worth mentioning the General Problem Solver (GPS). It was a program developed by Allen Newell and Herbert Simon between 1957 and 1959. GPS was an attempt to create a general solution to a wide range of problems by imitating human reasoning processes.

The main goal of GPS was to attempt to solve the problem of general intelligence —that is, to create a system capable of dealing with a variety of problems, not just specialized problems in a single field, as was the case with Logic Theorist. GPS could divide large, complex problems into smaller ones, and its main distinguishing feature was its ability to generalize. It could be applied to various problems and in many fields, not just specialized applications.

Perceptrons.

Perceptrons became a hot topic. They were one of the first models based on neural networks, designed in the 1950s and 1960s by Frank Rosenblatt. The structure of the human brain and attempts to model individual neurons inspired the perceptron concept.

In its basic form, a perceptron accepted multiple inputs, analyzed them, and then generated a single output. Each input to the perceptron was weighted, meaning that each input had a value that influenced the final output. The perceptron summed the weighted inputs and then transformed them to produce an output. If the sum of the weighted inputs exceeded a certain threshold, the perceptron generated a positive output; otherwise, it produced a negative output.

Although perceptrons were criticized and their popularity declined, they ultimately became the basis for the subsequent development of deep learning algorithms. Modern neural networks, which underpin many advanced artificial intelligence systems, expand and extend the concepts introduced by perceptrons, possessing additional layers and more complex activation functions. However, it was the work on perceptrons that ushered in the era we now call machine learning.

Eliza.

In 1966, while working at the Massachusetts Institute of Technology (MIT), Joseph Weizenbaum created Eliza, a simple computer program that simulated conversation. It was the first example of natural language processing.

This program, however, was not "conscious" in any way, nor did it have the ability to evolve as modern AI solutions do. The main technique was keyword recognition and script-based response generation, and the main use was to simulate the work of a therapist.

Even though the program was quite simple, many people responded to it as if it truly understood their feelings, which surprised Weizenbaum himself.

SHRDLU.

SHRDLU, created by Terry Winograd in the 1960s and 1970s, is a human language processing tool that could interpret and execute commands in a fictional "block" environment. It could also describe, answer questions about, and manipulate various objects.

DENDRAL.

Created in 1965, this pioneering program in chemistry laid the foundation for many subsequent reasoning systems in various other fields. Its authors were Edward Feigenbaum and Joshua Lederberg at Stanford University in California.

MYCIN.

Developed in the 1970s at Stanford University, MYCIN was one of the first computer-based medical systems. It was designed to diagnose bacterial infections and recommend appropriate antibiotics.

AAAI.

In 1979, the American Association of Artificial Intelligence was founded and is now known as the Association for the Advancement of Artificial Intelligence (AAAI).

Artificial intelligence winter: a time of stagnation.

The second half of the 1970s saw the so-called first artificial intelligence winter. Overly high expectations for AI and insufficient computing power led to a reduction in research funding.

The second AI winter occurred between 1987 and 1993. This period of low interest from consumers, the public, and the private sector again led to a decline in research funding. Both private investors and the government lost interest in AI and withheld funding due to the high costs relative to the limited benefits.

Since the dawn of computers, scientists have continually strived to recreate the human intellect in a machine. From the Logic Theorist, the first attempts at computer-assisted proof of mathematical theories, to the first virtual "therapist."

And despite this period of stagnation, they continued to work on new technologies. Here are the most important ones created during this difficult period.

Neocognitron.

Introduced by Kunihiko Fukushima in 1980, it is the progenitor of convolutional neural networks, which became the foundation of modern deep learning.

Backpropagation.

This backpropagation error algorithm, popularized in the 1980s by Geoffrey Hinton, became the key to effectively training multilayer neural networks.

NetTalk.

Created by Terrence Sejnowski and Charles Rosenberg in 1987, this program learned to pronounce words just like humans, which was early evidence of the possibilities of deep learning.

The turn of the 1990s: intelligent Deep Blue vs. world chess champion.

In 1997, IBM's Deep Blue computer defeated Garry Kasparov, the world chess champion. This was a milestone in the history of AI, demonstrating that machine intelligence could rival humans in solving complex intellectual tasks.

That same year, Windows first introduced human speech recognition software.

AI in the early 21st century: from machine learning to deep learning.

The beginning of the 21st century was another period of prosperity and enormous acceleration in development. It was also a time when entirely new technological concepts emerged.

The development of machine learning.

Since the 2000s, machine learning has become a key element of AI. With access to large amounts of data and immense computing power, machines have begun to learn quickly and automatically optimize their actions.

Deep learning.

Technologies like neural networks have allowed AI to recognize images, process speech, and predict trends with unprecedented accuracy.

Watson.

IBM Watson is an AI system that defeated the champions on the game show "Jeopardy!" in 2011, demonstrating AI's potential in natural language processing and general understanding.

Siri.

In 2011, Apple introduced Siri, the first popular virtual assistant.

AlexNet.

In 2012, a network designed by Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton achieved breakthrough results in the ImageNet image recognition competition, accelerating the development of deep learning.

AlphaGo.

In 2016, AlphaGo, a program created by DeepMind, defeated the world champion in the game of Go, considered more complex than chess.

All of the above-mentioned projects, along with many others, laid the foundations for modern AI technologies and were subsequent turning points in its evolution.

The path of AI development was full of twists, doubts, and skepticism, but step by step it led us to today's advanced systems, such as ChatGPT.

Modern natural language processing models: ChatGPT and others.

The development of artificial intelligence has always had one primary goal: fully utilizing natural language. To achieve this, grammatical rules and linguistic approaches, statistical methods, or the aforementioned neural networks (e.g., LSTM) have been used so far.

While the most advanced solutions were able to find contextual connections between words on a broad scale, they still failed to grasp the meaning of longer utterances.

A real breakthrough came in 2017 with the publication of the famous paper "Attention is All You Need," which described the so-called transformer mechanism.

This proved crucial for understanding longer utterances. Building on the transformer architecture, increasingly larger networks began to emerge, trained on increasingly larger datasets. One of the most important models utilizing the transformer architecture was the BERT (Bidirectional Encoder Representations from Transformers) model.

The next step in development is the development of a model called the Generative Pre-trained Transformer.

In 2020, OpenAI began beta testing a version of GPT-3 that could already generate content (i.e., write text, create poetry, or create computer code). While it wasn't the first model of its kind, it was the first whose content was nearly indistinguishable from that created by humans. Such models are characterized by a very large number of parameters, hence their general name, Large Language Models (LLMs).

GPT-3, the predecessor of the version used in the latest ChatGPT 4 release, has 175 billion parameters. For comparison, the previously mentioned BERT used 340 million parameters.

In November 2022, another breakthrough occurred. OpenAI publicly released ChatGPT 3.5. The first million users registered within five days of launch. In just two months, the number of users reached 100 million.

In November 2023, Microsoft announced the launch of Copilot for Office 365. This integrated assistant supports users in everyday office tasks by automatically generating content and analyzing data.

In December 2023, Google launched Gemini 1.5, the latest version of its AI model, combining chatbot technology with multimodal image and text processing.

In April 2024, Meta AI announced LLaMA 3, designed for faster and more energy-efficient data analysis at scale.

It's also worth noting that in August 2024, a team of Polish researchers will introduce the first highly developed Polish language model, called Bielik. It is adapted to the specifics of the Polish language and capable of analyzing and generating texts like the latest global models.

Intelligent Robots.

Here is a list of the most interesting events driving the progress of robotics:

• 1929: Japanese professor Makoto Nishimura built the first Japanese robot called Gakutensoku.

• 1961: The first industrial robot, Unimate, began working on an assembly line at General Motors in New Jersey, transporting castings and welding car parts (which was deemed too dangerous for humans).

• 2000: Professor Cynthia Breazeal developed the first robot that could simulate human emotions using the face, including eyes, eyebrows, ears, and mouth. It was called Kismet.

• 2002: The first Roomba robot was launched.

• 2003: NASA sent two rovers to Mars (Spirit and Opportunity), which moved on the planet's surface without human intervention.

• 2016: Hanson Robotics created a humanoid named Sophia, who was the first to receive formal citizenship in 2017. Sophia was the first human-like robot able to see, emote, and communicate.

• 2017: XiaoYi robot passed the Chinese medical examination, becoming the first artificial intelligence to achieve such a result.

Autonomous vehicles.

• 1961: James L. Adams created The Stanford Cart, which became one of the pioneering prototypes of an autonomous vehicle. In 1979, it successfully navigated a room full of chairs without human intervention.

• 1986: Ernst Dickmann and his team at the Bundeswehr University of Munich created and demonstrated the first driverless car (or robot car). It could travel at speeds of up to 55 miles per hour on roads that had no other obstacles or moving vehicles.

• 1997: The Japanese built a prototype of the first-ever unmanned passenger vehicle, incorporating cameras that fed data to a computer to process images of the road. It could reach speeds in excess of 30 km/h.

• 2005: The first fully autonomous car crossed the desert. It can still be seen today at the Smithsonian Institution National Museum of American History.

• 2014: Tesla Motors demonstrates the first version of Autopilot. The Model S is capable of lane control, braking, and adjusting to speed limits according to road signs. It also features self-parking and automatic software updates.

• 2018: Waymo launched the first fully driverless taxi service in Phoenix, USA.

• 2021: Hyundai's Ioniq 5 Robotaxi achieves Level 4 autonomy (the highest defined level is 5). The vehicle is expected to be used in public transport from 2023.

📌 Artificial intelligence applications in the future: what awaits us?

The future of AI brings many opportunities, but also challenges related to ethics and security. Therefore, one of the key issues for further development will be the legal regulation of all aspects of this extremely dynamic field.

Areas where we can expect rapid progress are:

• The development of so-called Artificial General Intelligence. AGI is capable of performing any intellectual task.

• Application of AI in scientific research and space exploration.

• Increasingly sophisticated and effective solutions for medicine and healthcare.

• Improving autonomous systems in transport and robotics.

We hope this short history of artificial intelligence has met your expectations. Finally, we invite you to watch the video, which provides a fascinating explanation of how bots were created.

See also other related articles:

• What Noam Chomsky thinks about ChatGPT.

• A brief history of coaching - from the 19th century to the present day.

About the author:

Ryszard Skarbek is a certified business coach, mentor, and HR advisor. He has over 30 years of professional experience, including 22 years in an international corporation operating on six continents.

As a coach and mentor, Ryszard works mainly with middle and senior managers and business owners.

His main specializations are 21st-century leadership, interpersonal communication, self-awareness, HR, IT, and Project Management.