Chapter 2, Part 1 — AI and Its History

Chapter 2, Part 1 — AI and Its History

Artificial Intelligence (AI), once a concept confined to the realm of science fiction, has almost overnight evolved into one of the most dynamic and rapidly growing technologies the world has ever seen. The meteoric surge in AI’s popularity is evident in the rise of generative consumer-facing AI platforms like ChatGPT. To give an idea of how popular this AI software has become, the search term “ChatGPT” currently ranks as the 17th most searched term this year, sandwiched between the terms “Google” and “porn,” highlighting AI’s growing presence in the public consciousness (Barenholtz, 2023). This remarkable growth has emerged in less than a year, showcasing just how much AI’s influence has grown and will likely continue to grow.

In a paper published in 2020, Chatterjee and Bhattacharjee defined AI as a “computing system capable of engaging in human-like processes such as adapting, learning, synthesizing, correcting and using various data required for processing complex tasks” (Chatterjee & Bhattacharjee, 2020). As the name Artificial Intelligence suggests, at its core, AI is about simulating and expanding human intelligence through machine and computer systems.

While AI machines are capable of tasks typically requiring human intelligence, such as learning, reasoning, and problem-solving, AI programs are far from sentient. In other words, the AI does not ‘think’ in the same way humans do; instead, it relies on processing mass amounts of data to produce a desired output. Most AI systems, unlike humans, can’t learn and perform tasks simultaneously so their processes are typically split into two phases: training and inference, which essentially refers to the process of using a trained AI model to make predictions or decisions on new, unseen data. Therefore, unlike humans, the AI does not come to conclusions based on reasoning but instead arrives at answers based on the models that were created for them, and, in some cases, they created themselves (Kissinger et al., 2022, p. 64).

AI Key Terms

While many aspects of AI can be categorized in different ways with many technical considerations, for this thesis, the main components can be categorized into three primary components: 1) Foundational Technologies, 2) Application-specific Technologies, and 3) Output-based Technologies as follows:

Foundational Technologies form the essential building blocks for AI systems. These technologies include computing power, data storage, algorithms, and networking technologies and provide the infrastructure for training and running AI models, processing massive datasets, and connecting AI systems in real-world applications. Foundational technologies include “machine learning”, a foundational, and early application of AI, that uses algorithms to enable computers to perform tasks without the need to program each task individually. In other words, the computer actually learns new tasks as it is being used and gets “smarter” with each subsequent use.

Most of us are already familiar with some form of Foundational AI. Applications that use Foundational AI —which we already use today — include Natural Language Processing and Large Language Models. Natural Language Processing (NLP) allows computers to understand, interpret, and respond to human language in a usable way such as airline ticketing AI “agents” who allow you to book your flight entirely through a computer that sounds like a human but focuses on very specific predictable tasks. Large Language Models (LLM) advance this technology even more by using large datasets of voices that literally “train” AI systems, to process, generate, and understand human language at an even larger and more “realistic” scale, and with greater nuance for broader application (Kissinger et al., 2022, p. 64).

Application-specific technologies are tailored to solve particular problems or operate within specific domains. They utilize foundational technologies—such as computing power, data storage, algorithms, and networking—to create AI solutions that are finely tuned for specific fields like healthcare, finance, and autonomous driving. This customization allows the AI models to meet the unique demands and complexities of each sector effectively.

Output-based technologies, often referred to as “generative AI,” are the newest form of AI. These technologies enable AI systems to actually create new content such as text, images, code, or music. Generative AI leverages machine learning, natural language processing, and large language models to generate entirely new content such as ChatGPT. Generative AI stands out as a particularly impactful area for the future and is positioned to revolutionize creative industries although raising significant concerns regarding authenticity and intellectual property rights. To work, Generative AI is built on “Deep Learning” – the most advanced subset of AI—which seeks to simulate the neural network of the brain by creating multiple layers of analysis to experience and understand the incoming data. These layers are largely unknown and offer little transparency as to how the machine got to its conclusions. This is why advances in AI, especially generative AI, can be problematic in fields like education, where understanding the intricacies of learning is vital.

Categorizations of AI Intelligence

Already, AI is surpassing human capabilities to perform in varying degrees, in various ways, and for various purposes:

Narrow AI specializes in a single area, surpassing human capabilities in a single domain such as playing chess or analyzing stock markets (Hamilton et al., 2023).

Artificial General Intelligence (AGI) functions at a more “human” level and is capable of managing a wide range of tasks, matching and even surpassing human ability in complex and cognitive activities such as law, medicine, policy analysis, research, writing, and decision-making (Hamilton et al., 2023).

Artificial Super Intelligence (ASI), by far the most advanced, is already demonstrating capabilities far beyond any human level in all cognitively demanding tasks. Ray Kurzweil, inventor, machine learning pioneer, and futurist, posits that ASI could potentially exceed the collective cognitive abilities of every human who has ever existed (Hamilton et al., 2023).

History of AI

Early Beginnings
Humans have imagined notions of “artificial intelligence” for thousands of years with origins of this idea seen in ancient mythologies, such as in Greek mythology where the god Hephaestus crafted robots capable of human-like tasks, such as Talos, a mythical automaton designed to guard the shores of Crete (Kissinger et al., 2022, p. 59). This fascination with mechanized assistants was not just limited to mythology. In the Han Dynasty of China, there is evidence of a mechanical orchestra built to entertain the emperor (Encyclopaedia Britannica, 2023). These early visions of intelligent machines laid a conceptual foundation for modern AI. However, it was the vast increase in computing power following World War II, combined with advancements in algorithms and the growing availability of data, that formulated the AI as we know it today.

The contemporary field of artificial intelligence truly began to take shape in the 20th century, largely thanks to American computer scientist John McCarthy who coined the term “artificial intelligence” during the pivotal Dartmouth Summer Research Project on Artificial Intelligence in 1956. This seminal conference sparked two decades of prolific research in the field (Stone et al., 2016). From 1957 to 1974, AI research flourished, driven by a steady increase in computing power and government funding due to the field’s early successes and the advocacy of leading researchers (Anyoha, 2017).

In fact, during this period, enthusiasm for the promise of AI was so high that policymakers also took the prospect extremely seriously, with Lyndon B. Johnson establishing the US National Commission on Technology, Automation, and Economic Progress in 1964 (Hamilton et al., 2023).

In education, a significant development occurred in 1964 with Joseph Weizenbaum’s creation of an early language processing model which was able to mimic and simulate human conversation patterns. Building on this advancement, Jaime Carbonell developed SCHOLAR, the first student-oriented instructional program that offered instant feedback in natural language on the quality of students’ responses. SCHOLAR would later be known as an Intelligent Tutoring System (ITS), representing a key intersection of AI and educational technology, and is a popular idea even to this day (Guan et al., 2020).

In the mid-1970s and early 1980s, enthusiasm and funding for AI research declined because the technology had not yet advanced enough to process the massive amounts of data required for creating powerful AI models. Hans Moravec, a doctoral student under John McCarthy at the time, captured the technological limitations of that era, saying that “computers were still millions of times too weak to exhibit intelligence” (Anyoha, 2017). During this period, the U.S. Congress also criticized the substantial spending on AI research (Haenlein & Kaplan, 2019).

A significant shift in how AI was perceived occurred in the early 1980s, spurred by an increase in funding, particularly from the Japanese government through their “Fifth Generation Computer Project.” It was during this time that American scientists John Hopfield and David Rumelhart introduced innovative “deep learning” techniques. These techniques revolutionized AI by enabling computers to learn from experience rather than relying solely on more traditional statistical methods (Anyoha, 2017). This approach to machine learning aimed to replicate how humans learn, through repeated exposure and iterative adjustment, rather than merely following explicit programming. This opened the door to the fundamental thinking that undergirds AI today.

The Recent AI Boom
While the economic challenges of the late 1980s led to a temporary slowdown in AI development, the subsequent decades saw a resurgence (Anyoha, 2017). Despite minimal funding from the US government and limited public enthusiasm during the 1990s and 2000s, AI experienced significant growth. This resurgence was fueled by increased computational power, which facilitated the development of neural networks (Mukherji, 2020). One of the most notable milestones in the advancement of artificial intelligence occurred in 1997 when IBM’s Deep Blue, a sophisticated chess-playing computer program, famously defeated Gary Kasparov, the world chess champion and grandmaster at the time (Anyoha, 2017).

Influenced by the work of John Hopfield and David Rumelhart, renowned for their research on neural networks, the focus in the AI field began to shift. Rather than working to program machines to mimic human intelligence, the emphasis moved towards enabling machines to learn on their own, marking the beginning of the machine learning era (Kissinger et al., 2022, p. 59). This machine learning era was further driven by the internet’s ability to collect vast quantities of data, coupled with the increased availability of computational power and storage for processing this data, AI models became increasingly powerful. Milestones such as NASA’s autonomous Mars rover in 2003 and the launch of Apple’s Siri in 2011 highlighted the progress in AI technology (Anyoha, 2017).

In 2017, the documentary AlphaGo showcased a pivotal moment in artificial intelligence, capturing how an AI computer successfully defeated Lee Sedol, the top-ranked Go player (Kohs, 2017). This event not only stunned Go enthusiasts worldwide but also underscored AI's emerging power. Since then, AI has evolved from a theoretical concept to a widely recognized reality, as demonstrated by platforms such as ChatGPT, which millions around the world now use.

The current capabilities of artificial intelligence, which range from generating creative content to interpreting and processing human language, signify a remarkable advance in technology. This progress holds the potential to profoundly influence our future, especially in the realm of education—potentially transforming how we teach, learn, and evaluate knowledge.

References

Chen, C. (2023, May 9). AI will transform teaching and learning. Let’s get it right. Stanford HAI. https://hai.stanford.edu/news/ai-will-transform-teaching-and-learning-lets-get-it-right

Hamilton, A., Wiliam, D., & Hattie, J. (2023). The future of AI in education: 13 things we can do to minimize the damage [Working paper]. ResearchGate. https://www.researchgate.net/publication/373108877

Holmes, W. (2023). AIED—Coming of age? International Journal of Artificial Intelligence in Education, 33(1), 1–6. https://doi.org/10.1007/s40593-023-00333-4

Kohs, G. (Director). (2017). AlphaGo [Film]. Moxie Pictures.

Lynch, S. (2023, April 3). AI benchmarks hit saturation. Stanford HAI. https://hai.stanford.edu/news/ai-benchmarks-hit-saturation

Perrault, R., Clark, J., & AI Index Steering Committee. (2024). AI Index Report 2024 – Artificial Intelligence Index. Stanford Institute for Human-Centered Artificial Intelligence. https://aiindex.stanford.edu/report/