Artificial Intelligence (AI) has ‘burst’ into our daily lives, becoming an integral part of it. Education, entertainment, work, and research – AI in the form of neural networks is used everywhere. Yes, this is far from the artificial intelligence described by science fiction writers – with self-awareness and instant evolution, acquisition of emotions and desires. But the future holds further possibilities. For now, AI constitutes a computer program, and it’s useful to explore its history. All the more so because the history of artificial intelligence didn’t start recently, and not even in the 21st century.
Where It All Began
The year 1943 is considered the beginning of the artificial intelligence history. In that year, Warren McCulloch and Walter Pitts published an article detailing a mathematical model of a neural network. In 1949, neuropsychologist Donald Hebb published a book that described a theory of building and developing neural pathways in the brain.
Thus, the history of the creation of artificial intelligence began with two seemingly disparate theories: a mathematical model and discoveries in neurobiology.
In the 1950s, theory turned to practice:
- Alan Turing published his celebrated test, determining the intelligence of a machine;
- Dean Edmonds and Marvin Minsky created a computer with a neural network (SNARC);
- Isaac Asimov published his famous “Three Laws of Robotics.”
In 1956 the concept of “Artificial Intelligence” was coined, and two years later, John McCarthy developed the AI programming language Lisp. The first artificial intelligence project was founded in 1959 by John McCarthy and Marvin Minsky at MIT.
The beginning of AI’s history was turbulent and full of promises. New programming languages, programs, and algorithms were developed, but by the early 1970s, a clear lack of progress tied to insufficient computing power heralded the first “AI winter.” The first branch of artificial intelligence’s evolution turned out to be a dead end.
Further Development
The “thaw” began in 1980 with the development of R1 – the first successful expert system. A focus on expert systems and the development of capabilities for them came from the USA and Japan. There even emerged a market for Lisp machines – computers created for working with the AI language Lisp. But by the early ’90s, with the emergence of cheaper machines, the market collapsed. The second “AI winter,” another dead-end branch of development.
Yet, in the ’90s, a new concept emerged, the ”intelligent agent,” a system that can autonomously complete a task assigned by a human. Confirmation of this new direction’s validity came when a computer defeated a human: in 1997, the IBM computer Deep Blue won in a chess match against world champion Garry Kasparov.
The most vigorous development began in the 2000s. More powerful computers, the internet, mobile technologies – all required new solutions for processing vast amounts of data. And those solutions were found. Today, we are users of those research outcomes.
AI Evolution
Modern neural networks seem incredibly smart and capable of anything. They might soon gain self-awareness, desire world domination, or start a global war. But – no. Modern AI is incapable of this because, for all its capabilities, it is a descendant of T9.
It was the technology scientists call a “language model” that became the great-great-grandparent of the neural networks we know today. Briefly, the evolutionary process looks like this:
- T9 in the ’90s – guesses which letter is next when entering a word;
- T9 in the 2010s – chooses words considering context, punctuation, auto-correction;
- 2020s – ChatGPT can write a TV manual in verse.
The evolution is rapid, but the operating principle of T9 and even ChatGPT remains unchanged. It’s language modeling, or predicting the next word considering the context. And the more data the neural network has, the more capabilities it possesses.
Countries Involved in AI Development
The history of artificial intelligence development began in the USA, and every stage of AI evolution took place there as well. It was in this country where the first steps in the field of machine learning were taken, and extensive research programs were developed.
In the ’80s, Japan made its foray into this realm, but its programs turned out to be ineffective.
Companies Engaged in AI Development
Naturally, companies connected with computer technologies participated in machine learning research and development:
- IBM
- Digital Equipment Corporation
- Symbolics
- Lisp Machines Inc.
- Boston Dynamics
- iRobot
- Apple
- Amazon
- Hanson Robotics
- Waymo
- Baidu
History of Artificial Intelligence in Russia
One should not think that the history of artificial intelligence in Russia started only in recent years. Russia, too, had its own developments. In 1962, in Zelenograd, a machine capable of calculating the trajectories of flying objects and predicting other events was developed. Thus, as early as the ’60s, the USSR was working on machine learning. Victor Glushkov made a significant contribution in this field, and many of his developments are still in use today.
If we study the history of AI in Russia, from cumbersome machines on punch cards to modern neural networks, we’ll get this chronology:
- 1945 – 1964 – creation of individual programs for solving logical tasks. The ALPEV LOMI program for automatic theorem proving.
- 1954 – The “Automata and Thinking” Seminar. The beginning of AI history with a division into neurocybernetics and “black box” cybernetics (informatics).
- 1960 – development of the “Kora” algorithm for modeling brain activity and pattern recognition (M. Bongard).
- 1965 – 1980 – development of science of situational management under the leadership of D.A. Pospelov.
- 1980-1990 – development of knowledge representation languages (machine learning), creation of the REFAL language.
- 1988 – foundation of the Artificial Intelligence Association.
AI development in the USSR was no less successful than in the USA. Until the ’70s – from 1975 onwards, technological lag began to take its toll. Today, Russia lags behind Western countries by 5-7 years, so new developments are based on research and work done in the USA. But this applies to neural networks, AI in the form we know it now.
The Future of AI
Analysts and futurists have differing forecasts. But, most likely, the AI branch based on language modeling will not evolve qualitatively differently. Neural networks will become smarter and more efficient, but they will remain within their bounds.
A qualitative transition from a simple program to full-fledged AI with self-awareness similar to human is only possible with the use of quantum technologies and bionic systems, discoveries in other directions of science. The first steps in this direction have already been taken, including in Russia. What the history of AI development in the future will be is something only time will tell.
