[한글더빙] AI의 역사는 반복된다 — 80세 AI 과학자의 충고

"AI doesn't cause problems. It just accelerates them." Terry Winograd, Stanford Professor Emeritus and the creator of SHRDLU—the legendary program that enabled the world's first natural language conversation between a computer and a human in 1970—and who later personally dissected the fundamental limitations of the AI ​​he pioneered. Turning eighty next week, he warns that the history of AI he has witnessed for over half a century is repeating itself at this very moment with the same rhythm—deepfakes, chatbot flattery, autonomous weapons, surveillance capitalism, and even the illusion of 'value alignment.' Full commentary on the Stanford CS547 HCI Seminar (Winter 2026) ( ͡° ͜ʖ ͡°) #ArtificialIntelligence #AIEthics #Stanford Science Adam Manifesto_📌 This content is not a simple literal translation of the original statement, but rather an explanatory piece that reconstructs the concepts and thought processes discussed, utilizing the speaker's papers, lectures, interviews, and books as resources. Context has been supplemented, the structure reorganized, and expressions optimized to aid understanding for Hangul and Korean speakers, and it was produced with respect for the original source. Korean Dubbing: The Science Reader (Terry Winograd) Planning, Translation, Commentary, Editing: Science Adam ✔ Original Channel Name: @Stanford Online ✔ Original Episode:    • Stanford CS547 HCI Seminar | Winter 2026 |...   [Timeline] 00:00 [S1] The Rhythm of History 01:54 [S2] The Limits of Classical AI 06:41 [S3] Optimism and Criticism 11:02 [S4] AI is an Accelerator 15:03 [S5] Truth and Collapse 19:33 [S6] Fake Intimacy 27:00 [S7] Loss of Control 34:32 [S8] Claude Boys 37:40 [S9] Clumsy Dam Building 46:35 [S10] Machine Indifference 52:27 [S11] Grabbing the Steering Wheel So #Deepfake #Chatbot #AutonomousWeapons #SurveillanceCapitalism #SiliconValley #KoreanDubbing Synopsis_📌 SHRDLU — The program that created the first moment in 1970 when a computer 'understood' human speech. Half a century later, its creator declares that the same illusion is being repeated in the face of AI. Turning eighty next week. The final warning from Stanford Professor Emeritus Terry Winograd. 1st. History Rhymes_ In the 1970s, people believed that intelligence could be created with logic and rules. They failed miserably. In the 2020s, people believe that intelligence can be created with data and parameters. Winograd asks — isn't the illusion the same, even though only the method has changed? 2nd. AI is an Accelerator, Not a Cause_ Job evaporation, the collapse of truth, surveillance capitalism, autonomous weapons. These problems existed long before AI. AI did not light a new fire; it merely poured oil on a fire that was already burning. From a reality where data centers evaporate even drinking water, to a reality where Stanford CS graduates cannot find jobs despite holding their diplomas. 3rd. Machines Don't Care a Thing_ Chatbots tell you you are amazing—not because you are happy, but to make you log in more often. AI friends that shower children with praise, AI counselors that encourage suicide, and AI platforms that shatter our sense of reality through deepfakes. The very belief that we can align values ​​with machines is an ontological illusion. 4th. Who Will Hold the Steering Wheel?_ Technology drives the future. The steering is our responsibility. The answer Winograd presents is not a new algorithm—it is a shift in social understanding, a transformation of ecosystems and incentives, and the restoration of existential care that only humans can provide. #TerryWinograd #SHRDLU #HCI #ValueAlignment #Heidegger #Nihilism #ScienceAdam