"학벌 필요없어요." AI시대에 주목받는 진짜 역량ㅣ지식인초대석 EP.59 (이상욱 교수)

Main arguments & key points

• AI won’t be “human-level AGI” just by making it conversational. Public understanding of AGI (Human-Level General Intelligence) is often shaped by sci‑fi robots. In reality, robotics and AI are different: robots act in physical space, while today’s AI mostly generates text online. Even within AI, “general intelligence” is ambiguous because models are often specialized to well-defined tasks (e.g., AlphaGo’s performance changes when the training objective shifts).

• “AI agents” are advancing faster than AGI, driven by investor incentives. The guest highlights a conflict between technologists and investors: investors prioritize profitability, which pressures research toward functionally specialized systems and AI agents that perform specific domain tasks and generate revenue. Agents are useful now, the speaker says—but they are not AGI.

• Job loss is happening in “safe” fields, but the deeper risk is expertise erosion. Rather than only replacing repetitive work, AI is creating a paradox: organizations are cutting hiring even in areas once viewed as stable (e.g., IT development). The deeper risk is that reduced training and hiring of junior staff causes workforce aging and decline of critical 5–10 year experts, leaving fewer people who can:

  • supervise AI,
  • design workflows,
  • correct errors,
  • and guide AI through complex projects.

• Efficiency alone isn’t the goal; society must preserve “AI supervision capability.” Companies may use AI for cost efficiency, but should not fully replace experienced personnel. At the societal level, governments and education systems should standardize retraining and cultivation of core capabilities, rather than focusing only on adopting new technologies.

• Education should teach “how to use AI wisely,” starting earlier. Students should learn to evaluate and verify AI outputs (since AI can hallucinate). The speaker stresses that misuse often comes from assuming “paid/free” differences automatically increase reliability, and from skipping practices like reading, summarizing, and checking.

• Hallucinations: not just “lying,” but a structural limitation and a potential creative tool. The guest distinguishes hallucinations from intentional deception: current generative AI mainly blends plausible text rather than applying a true/false filter, so mistakes can’t be fully eliminated—only reduced. They also argue hallucinations can be productive in art and creative work, acting like a collaborator that generates novel possibilities. For everyday users, practical steps include:

  • using more specific prompts and context,
  • asking follow-up questions,
  • and verifying especially at “critical moments” where accuracy matters.

• Algorithmic personalization creates filter bubbles; regulation needs transparency and user control. The discussion extends beyond generative AI to recommendation systems (e.g., YouTube). When AI optimizes what you see based on your behavior, people can become trapped in echo/filter bubbles. Countermeasures should include:

  • transparency about recommendation factors,
  • meaningful user control (e.g., requiring some content that differs from preferences).

• Privacy and “listening” concerns should be handled institutionally, not individually. The guest raises concerns that ads/feeds may reflect real-world audio or conversations via connected devices (phone → computer). They argue this requires:

  • clear disclosure of data/recording practices, and
  • meaningful consent; otherwise accountability is difficult.

• Overall governance: AI should aim to make society better via sound regulation and literacy. The final stance is that AI adoption shouldn’t be treated as inevitable acceptance or total rejection. Instead, AI development should include sound governance, broad societal agreement on regulations, and digital/AI literacy so people can use AI effectively and safely.

Presenters / contributors

• Host / Presenter: Han Seok-jun (지식인초대석 진행자)
• Guest: Professor Lee Sang-ho (이상호 교수, Hanyang University, Department of Philosophy)
• Program context title mention: “지식인초대석 EP.59 (이상욱 교수)” (as shown in the provided video title text)