Trend AnalysisHistory & Area Studies
History of Computing and Early AI Development: From Punch Cards to Prompt Engines
The current AI revolution did not emerge from a vacuum. It rests on eight decades of conceptual breakthroughs, engineering feats, and institutional decisions whose consequences are still unfolding. Un...
By Sean K.S. Shin
This blog summarizes research trends based on published paper abstracts. Specific numbers or findings may contain inaccuracies. For scholarly rigor, always consult the original papers cited in each post.
Why It Matters
The current AI revolution did not emerge from a vacuum. It rests on eight decades of conceptual breakthroughs, engineering feats, and institutional decisions whose consequences are still unfolding. Understanding this history matters because the assumptions embedded in early computing, from Turing's formalization of computation to the Dartmouth Conference's framing of AI as symbol manipulation, continue to shape how we build and regulate intelligent systems today.
The history of computing also reveals recurring patterns of hype and disillusionment (the "AI winters" of the 1970s and late 1980s) that illuminate current debates about whether large language models represent genuine progress toward general intelligence or another bubble destined to deflate. Historical awareness is the best inoculation against both uncritical techno-optimism and premature pessimism.
Moreover, the social history of computing, who was included and excluded, whose problems were prioritized, how military funding shaped research agendas, provides essential context for understanding the biases and power structures embedded in contemporary AI systems.
The Science
Chatbot Evolution
Stone (2025) produced a comprehensive review spanning from ELIZA (1966) through Siri and Alexa to ChatGPT, tracing how chatbot technology evolved from simple pattern-matching scripts to transformer-based generative models. With 45 citations, the paper has become a widely referenced timeline of conversational AI development, documenting the interplay between academic research, commercial deployment, and shifting user expectations.
AI and Learning Sciences
Al-Amin et al. (2024) examined the shared history of AI and the learning sciences, arguing that from their earliest days both fields have been entangled: AI researchers borrowed theories of human cognition to build machines, while education researchers adopted AI tools to study learning. The paper traces this reciprocal relationship from Skinner's teaching machines through intelligent tutoring systems to GPT-powered educational assistants.
From Antiquity to the Present
Stone (2025) traced AI's conceptual prehistory from ancient automata and philosophical debates about mind-body dualism through Babbage's Analytical Engine, Turing's theoretical work, the Dartmouth Yevstratov (2025), expert systems, neural network winters, and the deep learning revolution. The synthesis emphasizes that AI's development was never linear but followed a punctuated trajectory driven by hardware advances, data availability, and theoretical insights.
The Analogue Computing Blind Spot
Rosen (2024) challenged the dominant digital-centric narrative of computer history by recovering the role of analogue computers in early computer art. The paper argues that historiographic bias toward digital machines has obscured a rich tradition of analogue computing that shaped creative practices and offered alternative computational paradigms.
Key Milestones in Computing and AI History
<
| Year | Milestone | Significance |
|---|
| 1936 | Turing's "On Computable Numbers" | Theoretical foundation of computation |
| 1943 | McCulloch-Pitts neural model | First mathematical model of neurons |
| 1945 | ENIAC operational | First general-purpose electronic computer |
| 1956 | Dartmouth Conference | AI named as a field |
| 1966 | ELIZA chatbot | First conversational interface |
| 1974-1980 | First AI Winter | Funding collapse after overpromising |
| 1986 | Backpropagation revival | Neural networks reborn |
| 1997 | Deep Blue beats Kasparov | Symbolic AI's landmark victory |
| 2012 | AlexNet wins ImageNet | Deep learning revolution begins |
| 2017 | Transformer architecture | Foundation for LLMs |
| 2022 | ChatGPT launch | Generative AI enters mainstream |
What To Watch
The history of computing is being rapidly rewritten as archives are digitized and oral histories collected from aging pioneers. Expect growing attention to non-Western computing histories, including Soviet cybernetics, Indian software industry origins, and Japanese fifth-generation computing projects, that challenge the Silicon Valley-centric narrative. The emergence of AI-generated history itself (LLMs producing historical narratives) will raise new historiographic questions about authorship, accuracy, and the feedback loops between AI systems and the histories written about them.
๋ฉด์ฑ
์กฐํญ: ์ด ๊ฒ์๋ฌผ์ ์ ๋ณด ์ ๊ณต์ ๋ชฉ์ ์ผ๋ก ํ ์ฐ๊ตฌ ๋ํฅ ๊ฐ์์ด๋ค. ํ์ ์ ์๋ฌผ์ ์ธ์ฉํ๊ธฐ ์ ์ ๊ตฌ์ฒด์ ์ธ ์ฐ๊ตฌ ๊ฒฐ๊ณผ, ํต๊ณ ๋ฐ ์ฃผ์ฅ์ ์๋ณธ ๋
ผ๋ฌธ์ ํตํด ๋ฐ๋์ ํ์ธํด์ผ ํ๋ค.
์ ์ค์ํ๊ฐ
ํ์ฌ์ AI ํ๋ช
์ ๊ฐ์๊ธฐ ๋ฑ์ฅํ ๊ฒ์ด ์๋๋ค. ์ด๋ 80๋
์ ๊ฑธ์น ๊ฐ๋
์ ๋ํ๊ตฌ, ๊ณตํ์ ์ฑ๊ณผ, ๊ทธ๋ฆฌ๊ณ ์์ง๋ ๊ทธ ๊ฒฐ๊ณผ๊ฐ ์ ๊ฐ๋๊ณ ์๋ ์ ๋์ ๊ฒฐ์ ๋ค ์์ ์ธ์์ ธ ์๋ค. ์ด ์ญ์ฌ๋ฅผ ์ดํดํ๋ ๊ฒ์ด ์ค์ํ ์ด์ ๋, ์ด๊ธฐ ์ปดํจํ
์ ๋ด์ฌ๋ ๊ฐ์ ๋ค, ์ฆ Turing์ ๊ณ์ฐ ํ์ํ์์๋ถํฐ AI๋ฅผ ๊ธฐํธ ์กฐ์์ผ๋ก ๊ท์ ํ Dartmouth Conference์ ํ๋ ์ด๋ฐ์ ์ด๋ฅด๊ธฐ๊น์ง, ์ด๊ฒ๋ค์ด ์ค๋๋ ์ฐ๋ฆฌ๊ฐ ์ง๋ฅํ ์์คํ
์ ๊ตฌ์ถํ๊ณ ๊ท์ ํ๋ ๋ฐฉ์์ ๊ณ์ํด์ ํ์ฑํ๊ณ ์๊ธฐ ๋๋ฌธ์ด๋ค.
์ปดํจํ
์ ์ญ์ฌ๋ ๋ํ ๊ณผ๋ ์ ์ ๊ณผ ํ๋ฉธ์ ๋ฐ๋ณต์ ์ธ ํจํด(1970๋
๋์ 1980๋
๋ ํ๋ฐ์ "AI ๊ฒจ์ธ")์ ๋ณด์ฌ์ฃผ๋๋ฐ, ์ด๋ ๋ํ ์ธ์ด ๋ชจ๋ธ์ด ๋ฒ์ฉ ์ธ๊ณต์ง๋ฅ์ ํฅํ ์ง์ ํ ์ง๋ณด๋ฅผ ๋ํ๋ด๋์ง ์๋๋ฉด ๊ฒฐ๊ตญ ๊บผ์ง ์ด๋ช
์ ๋ ๋ค๋ฅธ ๊ฑฐํ์ธ์ง์ ๊ดํ ํ์ฌ์ ๋
ผ์์ ์กฐ๋ช
ํ๋ค. ์ญ์ฌ์ ์ธ์์ ๋ฌด๋นํ์ ๊ธฐ์ ๋๊ด์ฃผ์์ ์ฃ๋ถ๋ฅธ ๋น๊ด์ฃผ์ ๋ชจ๋์ ๋ํ ์ต์ ์ ์๋ฐฉ์ฑ
์ด๋ค.
๋ ๋์๊ฐ, ์ปดํจํ
์ ์ฌํ์ฌ, ์ฆ ๋๊ฐ ํฌํจ๋๊ณ ๋ฐฐ์ ๋์๋์ง, ์ด๋ค ๋ฌธ์ ๊ฐ ์ฐ์ ์๋์๋์ง, ๊ตฐ์ฌ ์๊ธ์ด ์ฐ๊ตฌ ์์ ๋ฅผ ์ด๋ป๊ฒ ํ์ฑํ๋์ง๋ ํ๋ AI ์์คํ
์ ๋ด์ฌ๋ ํธํฅ๊ณผ ๊ถ๋ ฅ ๊ตฌ์กฐ๋ฅผ ์ดํดํ๋ ๋ฐ ํ์์ ์ธ ๋งฅ๋ฝ์ ์ ๊ณตํ๋ค.
๊ณผํ
์ฑ๋ด์ ์งํ
Stone (2025)์ ELIZA(1966)์์๋ถํฐ Siri, Alexa๋ฅผ ๊ฑฐ์ณ ChatGPT์ ์ด๋ฅด๋ ํฌ๊ด์ ์ธ ๋ฆฌ๋ทฐ๋ฅผ ์ ์ํ๋ฉฐ, ์ฑ๋ด ๊ธฐ์ ์ด ๋จ์ํ ํจํด ๋งค์นญ ์คํฌ๋ฆฝํธ์์ ํธ๋์คํฌ๋จธ ๊ธฐ๋ฐ ์์ฑ ๋ชจ๋ธ๋ก ์ด๋ป๊ฒ ์งํํ๋์ง๋ฅผ ์ถ์ ํ์๋ค. 45๊ฑด์ ์ธ์ฉ์ ๊ธฐ๋กํ ์ด ๋
ผ๋ฌธ์ ํ์ ์ฐ๊ตฌ, ์์
์ ๋ฐฐํฌ, ๊ทธ๋ฆฌ๊ณ ๋ณํํ๋ ์ฌ์ฉ์ ๊ธฐ๋ ๊ฐ์ ์ํธ์์ฉ์ ๊ธฐ๋กํ ๋ํํ AI ๋ฐ์ ์ ํ์๋ผ์ธ์ผ๋ก์ ๋๋ฆฌ ์ฐธ์กฐ๋๊ณ ์๋ค.
AI์ ํ์ต ๊ณผํ
Al-Amin et al. (2024)์ AI์ ํ์ต ๊ณผํ์ ๊ณต์ ๋ ์ญ์ฌ๋ฅผ ๊ณ ์ฐฐํ๋ฉฐ, ๋ ๋ถ์ผ๊ฐ ์ด์ฐฝ๊ธฐ๋ถํฐ ์๋ก ์ฝํ ์์๋ค๊ณ ์ฃผ์ฅํ์๋ค. AI ์ฐ๊ตฌ์๋ค์ ๊ธฐ๊ณ๋ฅผ ๊ตฌ์ถํ๊ธฐ ์ํด ์ธ๊ฐ ์ธ์ง ์ด๋ก ์ ์ฐจ์ฉํ์๊ณ , ๊ต์ก ์ฐ๊ตฌ์๋ค์ ํ์ต์ ์ฐ๊ตฌํ๊ธฐ ์ํด AI ๋๊ตฌ๋ฅผ ์ฑํํ์๋ค. ์ด ๋
ผ๋ฌธ์ Skinner์ ๊ต์ก ๊ธฐ๊ณ์์ ์ง๋ฅํ ํํฐ๋ง ์์คํ
์ ๊ฑฐ์ณ GPT ๊ธฐ๋ฐ ๊ต์ก ๋ณด์กฐ ๋๊ตฌ์ ์ด๋ฅด๋ ์ด ์ํธ์ ๊ด๊ณ๋ฅผ ์ถ์ ํ์๋ค.
๊ณ ๋์์ ํ์ฌ๊น์ง
Stone (2025)์ ๊ณ ๋์ ์๋ ์ฅ์น์ ์ฌ์ ์ด์๋ก ์ ๊ดํ ์ฒ ํ์ ๋
ผ์์์๋ถํฐ Babbage์ ํด์ ๊ธฐ๊ด, Turing์ ์ด๋ก ์ ์์
, Dartmouth Yevstratov (2025), ์ ๋ฌธ๊ฐ ์์คํ
, ์ ๊ฒฝ๋ง์ ๊ฒจ์ธ, ๊ทธ๋ฆฌ๊ณ ๋ฅ๋ฌ๋ ํ๋ช
์ ์ด๋ฅด๋ AI์ ๊ฐ๋
์ ์ ์ฌ(ๅๅฒ)๋ฅผ ์ถ์ ํ์๋ค. ์ด ์ข
ํฉ์ ์ฐ๊ตฌ๋ AI์ ๋ฐ์ ์ด ๊ฒฐ์ฝ ์ ํ์ ์ด์ง ์์์ผ๋ฉฐ, ํ๋์จ์ด ๋ฐ์ , ๋ฐ์ดํฐ ๊ฐ์ฉ์ฑ, ์ด๋ก ์ ํต์ฐฐ์ ์ํด ์ถ๋๋ ๋จ์์ (punctuated) ๊ถค์ ์ ๋ฐ๋์์ ๊ฐ์กฐํ๋ค.
์๋ ๋ก๊ทธ ์ปดํจํ
์ ์ฌ๊ฐ์ง๋
Rosen (2024)์ ์ด๊ธฐ ์ปดํจํฐ ์์ ์์ ์๋ ๋ก๊ทธ ์ปดํจํฐ์ ์ญํ ์ ๋ณต์ํจ์ผ๋ก์จ ์ปดํจํฐ ์ญ์ฌ์์ ์ง๋ฐฐ์ ์ธ ๋์งํธ ์ค์ฌ ์์ฌ์ ๋์ ํ์๋ค. ์ด ๋
ผ๋ฌธ์ ๋์งํธ ๊ธฐ๊ณ์ ๋ํ ์ญ์ฌํ์ ํธํฅ์ด ์ฐฝ์์ ์ค์ฒ์ ํ์ฑํ๊ณ ๋์์ ๊ณ์ฐ ํจ๋ฌ๋ค์์ ์ ์ํ๋ ํ๋ถํ ์๋ ๋ก๊ทธ ์ปดํจํ
์ ํต์ ๊ฐ๋ ค์๋ค๊ณ ์ฃผ์ฅํ๋ค.
์ปดํจํ
๋ฐ AI ์ญ์ฌ์ ์ฃผ์ ์ด์ ํ
<
| ์ฐ๋ | ์ด์ ํ | ์์ |
|---|
| 1936 | Turing์ "On Computable Numbers" | ๊ณ์ฐ์ ์ด๋ก ์ ํ ๋ |
| 1943 | McCulloch-Pitts ์ ๊ฒฝ ๋ชจ๋ธ | ์ต์ด์ ์ํ์ ๋ด๋ฐ ๋ชจ๋ธ |
| 1945 | ENIAC ์ด์ฉ ๊ฐ์ | ์ต์ด์ ๋ฒ์ฉ ์ ์ ์ปดํจํฐ |
| 1956 | Dartmouth Conference | AI๊ฐ ํ๋์ ๋ถ์ผ๋ก ๋ช
๋ช
๋จ |
| 1966 | ELIZA ์ฑ๋ด | ์ต์ด์ ๋ํํ ์ธํฐํ์ด์ค |
| 1974-1980 | ์ 1์ฐจ AI ๊ฒจ์ธ | ๊ณผ๋ํ ์ฝ์ ์ดํ ์๊ธ ์ง์ ๋ถ๊ดด |
| 1986 | ์ญ์ ํ ๋ถํ | ์ ๊ฒฝ๋ง์ ์ฌํ์ |
| 1997 | Deep Blue, Kasparov ๊ฒฉํ | ๊ธฐํธ AI์ ์ญ์ฌ์ ์น๋ฆฌ |
| 2012 | AlexNet, ImageNet ์ฐ์น | ๋ฅ๋ฌ๋ ํ๋ช
์ ์์ |
| 2017 | ํธ๋์คํฌ๋จธ ์ํคํ
์ฒ | LLM์ ํ ๋ |
| 2022 | ChatGPT ์ถ์ | ์์ฑํ AI๊ฐ ์ฃผ๋ฅ์ ์ง์
|
์ฃผ๋ชฉํ ์ฌํญ
์์นด์ด๋ธ๊ฐ ๋์งํธํ๋๊ณ ๋
ธ๋ นํ๋๋ ์ ๊ตฌ์๋ค๋ก๋ถํฐ ๊ตฌ์ ์ญ์ฌ๊ฐ ์์ง๋จ์ ๋ฐ๋ผ, ์ปดํจํ
์ ์ญ์ฌ๋ ๋น ๋ฅด๊ฒ ๋ค์ ์ฐ์ฌ์ง๊ณ ์๋ค. ์ค๋ฆฌ์ฝ ๋ฐธ๋ฆฌ ์ค์ฌ์ ์์ฌ์ ๋์ ํ๋ ์๋น์ํธ ์ฌ์ด๋ฒ๋คํฑ์ค, ์ธ๋ ์ํํธ์จ์ด ์ฐ์
์ ๊ธฐ์, ์ผ๋ณธ์ ์ 5์ธ๋ ์ปดํจํ
ํ๋ก์ ํธ๋ฅผ ํฌํจํ ๋น์๊ตฌ๊ถ ์ปดํจํ
์ญ์ฌ์ ๋ํ ๊ด์ฌ์ด ๋์์ง ๊ฒ์ผ๋ก ์์๋๋ค. AI๊ฐ ์์ฑํ ์ญ์ฌ ๊ทธ ์์ฒด(LLM์ด ์ญ์ฌ์ ์์ฌ๋ฅผ ์์ฑํ๋ ๊ฒ)์ ๋ฑ์ฅ์ ์ ์์ฑ, ์ ํ์ฑ, ๊ทธ๋ฆฌ๊ณ AI ์์คํ
๊ณผ ๊ทธ์ ๊ดํด ์ฐ์ฌ์ง ์ญ์ฌ ์ฌ์ด์ ํผ๋๋ฐฑ ๋ฃจํ์ ๊ดํ ์๋ก์ด ์ฌํ์ ์ง๋ฌธ์ ์ ๊ธฐํ ๊ฒ์ด๋ค.
References (4)
History of generative Artificial Intelligence (AI) chatbots: past, present, and future development.
Stone, J. (2025). From Punch Cards to Prompt Engines: The Shared History of Artificial Intelligence and the Learning Sciences. The Open/Technology in Education, Society, and Scholarship Association Journal, 4(3), 1-19.
Yevstratov, Y. (2025). THE HISTORY OF ARTIFICIAL INTELLIGENCE DEVELOPMENT: FROM ANTIQUITY TO THE PRESENT. ะััะฝะธะบ ะฝะฐัะบะธ ัะฐ ะพัะฒััะธ.
Rosen, M. (2024). In the beginning, there was the analogue. An alternative history of the beginnings of computer art. Artnodes, 0(34).