Artificial intelligence has become increasingly influential, but its conceptual foundations were established decades before the digital age.
To create a machine that works and thinks just like humans is a mid 20th-century idea that evolved over decades to reach its present stage. Just like any other machine or product, it has been through its ups and downs before finally bringing a huge change in the technology world. AI is the best example of science fiction becoming a reality.
The computer scientist Alan Turing was the first person to propose the idea of a thinking machine in his paper Computing Machinery and Intelligence in 1950. His work opened the door of research in AI for other people. In 1956, during the workshop of Dartmouth Summer Research Project, the term Artificial Intelligence was coined by the researchers. This event formally established AI as an academic field, attracting researchers to study it in subsequent decades.
After many years of research, a neural network called Perceptron, which could recognize patterns was developed by Frank Rosenblatt in 1957, and a chatbot known as Eliza capable of using natural language processing (NLP) was developed by Joseph Weizenbaum at MIT in 1966. Shakey the Robot was developed at SRI between 1966 and 1972 and could navigate autonomously and perform tasks using logical reasoning. This was the first phase of AI, so these systems were simple and could do only limited functions.
During the 1970s, AI research faced reduced funding and interest due to the limitations of early systems, a period often referred to as the first “AI winter.” In its initial phase, AI was not up to the expectation of people. Thus, the funding for AI research also decreased in that decade significantly.
Unfortunately, the breakthroughs in the 1980s were still not enough to attract more funding for research and to convince the world. The expectations that the world had from AI were not practical then, and the limited functions AI was performing were too costly. These challenges contributed to periods of reduced research activity, sometimes called “AI winters.”
Machine learning reached an improved level in the 1990s when AI models were on data sets. This advancement resulted in the development of Support Vector Machines, Decision Trees like bagging and boosting and Reinforcement Learning. SVMs and techniques like Reinforcement Learning made AI capable of solving complex problems, so the scope of AI in different industries also increased. It was being used for facial recognition, detection of fraud documents and document classification. From there onwards, AI kept on becoming more and more advanced with each passing year.
In the first two decades of the 21st century, many breakthroughs were achieved in deep learning, neural networks and training methods due to better computational power. Deep Belief Networks were developed by Geoffrey Hinton in 2006 to train deep neural networks without supervised learning. This showed others the way to enhance deep model learning. The introduction of the Transformer architecture in 2017 advanced NLP through self-attention mechanisms, enabling large language models (LLMs) such as GPT to generate text that resembles human writing.
Now LLMs (Large language models) are the base of AI in 2025. These models are trained on large datasets, allowing AI to respond to questions, mimicking human behaviors. LLMs can generate text that reflects patterns in language, producing outputs that often appear contextually appropriate, which is why they mesmerized the world. LLMs can produce outputs that may appear creative by combining learned patterns in novel ways. Though trained on large datasets, modern AI models, like ChatGPT, Dall-E etc., are able to generate creative text, image and video. Due to the convenience brought by these modern AI models, they are influencing every field of life, from writing to science. This aspect of AI has also worried creative people around the world.
The research is still ongoing in artificial intelligence. Researchers are now trying their best to reach the level of artificial general intelligence. At that level, AI would be able to do tasks that only intellectual humans are able to do, whether in art or science. Some researchers have also coined the term artificial superintelligence for the level beyond AGI. At that level, AI would be able to do those tasks which are still impossible for even humans. This last stage is theoretical, but judging from the evolution of AI over the decades and from what it can do today, it will not be a surprise if AI reaches the levels of AGI and ASI.
When it comes to the future of tech, AI’s calling the shots and running the show. AI is increasingly influential across various sectors, with potential benefits and challenges that continue to be studied and debated.
Image: DIW-Aigen
Read next:
• Study Finds Openness to AI’s Utility But Concern Grows Over Chatbots Replacing Real Human Relationships
• Industrial Age to Tech Age, The Changes and The Cycle of Innovation (infographic)
• New Report Shows that Energy Consumption by Data Centers is Going to Get Doubled by 2030
To create a machine that works and thinks just like humans is a mid 20th-century idea that evolved over decades to reach its present stage. Just like any other machine or product, it has been through its ups and downs before finally bringing a huge change in the technology world. AI is the best example of science fiction becoming a reality.
The computer scientist Alan Turing was the first person to propose the idea of a thinking machine in his paper Computing Machinery and Intelligence in 1950. His work opened the door of research in AI for other people. In 1956, during the workshop of Dartmouth Summer Research Project, the term Artificial Intelligence was coined by the researchers. This event formally established AI as an academic field, attracting researchers to study it in subsequent decades.
After many years of research, a neural network called Perceptron, which could recognize patterns was developed by Frank Rosenblatt in 1957, and a chatbot known as Eliza capable of using natural language processing (NLP) was developed by Joseph Weizenbaum at MIT in 1966. Shakey the Robot was developed at SRI between 1966 and 1972 and could navigate autonomously and perform tasks using logical reasoning. This was the first phase of AI, so these systems were simple and could do only limited functions.
During the 1970s, AI research faced reduced funding and interest due to the limitations of early systems, a period often referred to as the first “AI winter.” In its initial phase, AI was not up to the expectation of people. Thus, the funding for AI research also decreased in that decade significantly.
- Also read: Study Finds Public Can Detect AI Content Easily, But Acceptance Depends on Context and Purpose
Unfortunately, the breakthroughs in the 1980s were still not enough to attract more funding for research and to convince the world. The expectations that the world had from AI were not practical then, and the limited functions AI was performing were too costly. These challenges contributed to periods of reduced research activity, sometimes called “AI winters.”
Machine learning reached an improved level in the 1990s when AI models were on data sets. This advancement resulted in the development of Support Vector Machines, Decision Trees like bagging and boosting and Reinforcement Learning. SVMs and techniques like Reinforcement Learning made AI capable of solving complex problems, so the scope of AI in different industries also increased. It was being used for facial recognition, detection of fraud documents and document classification. From there onwards, AI kept on becoming more and more advanced with each passing year.
In the first two decades of the 21st century, many breakthroughs were achieved in deep learning, neural networks and training methods due to better computational power. Deep Belief Networks were developed by Geoffrey Hinton in 2006 to train deep neural networks without supervised learning. This showed others the way to enhance deep model learning. The introduction of the Transformer architecture in 2017 advanced NLP through self-attention mechanisms, enabling large language models (LLMs) such as GPT to generate text that resembles human writing.
Now LLMs (Large language models) are the base of AI in 2025. These models are trained on large datasets, allowing AI to respond to questions, mimicking human behaviors. LLMs can generate text that reflects patterns in language, producing outputs that often appear contextually appropriate, which is why they mesmerized the world. LLMs can produce outputs that may appear creative by combining learned patterns in novel ways. Though trained on large datasets, modern AI models, like ChatGPT, Dall-E etc., are able to generate creative text, image and video. Due to the convenience brought by these modern AI models, they are influencing every field of life, from writing to science. This aspect of AI has also worried creative people around the world.
The research is still ongoing in artificial intelligence. Researchers are now trying their best to reach the level of artificial general intelligence. At that level, AI would be able to do tasks that only intellectual humans are able to do, whether in art or science. Some researchers have also coined the term artificial superintelligence for the level beyond AGI. At that level, AI would be able to do those tasks which are still impossible for even humans. This last stage is theoretical, but judging from the evolution of AI over the decades and from what it can do today, it will not be a surprise if AI reaches the levels of AGI and ASI.
When it comes to the future of tech, AI’s calling the shots and running the show. AI is increasingly influential across various sectors, with potential benefits and challenges that continue to be studied and debated.
Read next:
• Study Finds Openness to AI’s Utility But Concern Grows Over Chatbots Replacing Real Human Relationships
• Industrial Age to Tech Age, The Changes and The Cycle of Innovation (infographic)
• New Report Shows that Energy Consumption by Data Centers is Going to Get Doubled by 2030
