{"@context":"https://neupai.io/schema/v0.2","@type":"StructuredNewsArticle","identity":{"article_id":"tech42_20260806_ai-scientist-role","canonical_url":"https://www.tech42.co.kr/%ed%98%84%ec%9e%a5-ai%ea%b0%80-%eb%b6%84%ec%84%9d%ea%b3%bc-%ea%b3%84%ec%82%b0%ec%9d%84-%eb%a7%a1%eb%8a%94-%ec%8b%9c%eb%8c%80-%ea%b3%bc%ed%95%99%ec%9e%90%eb%8a%94-%eb%ac%b4%ec%97%87%ec%9d%84/?utm_source=rss&utm_medium=rss&utm_campaign=%25ed%2598%2584%25ec%259e%25a5-ai%25ea%25b0%2580-%25eb%25b6%2584%25ec%2584%259d%25ea%25b3%25bc-%25ea%25b3%2584%25ec%2582%25b0%25ec%259d%2584-%25eb%25a7%25a1%25eb%258a%2594-%25ec%258b%259c%25eb%258c%2580-%25ea%25b3%25bc%25ed%2595%2599%25ec%259e%2590%25eb%258a%2594-%25eb%25ac%25b4%25ec%2597%2587%25ec%259d%2584","ai_url":null,"publisher":{"name":"테크42","domain":"www.tech42.co.kr","type":"online"},"author":"황정호 기자","published_at":"2026-08-06T22:39:45.000Z","updated_at":null,"language":"en","article_type":"analysis","originality":"self_produced"},"content":{"headline":"[On-site] In an era where AI handles analysis and computation, what should scientists do?","summary":"At a discussion held during the '2026 Korea AI Innovation Talent Connect,' panelists diagnosed that as AI automates the computation and analysis of research, the role of scientists is shifting toward question design, results verification, interpretation, and persuasion. Opinions were presented that, beyond simply using AI as a tool, the ability to identify AI's errors and prove industrial value based on principled understanding and domain knowledge is essential for future scientists.","topics":["AI","scientific research","talent development","drug development","astronomy","weather forecasting"],"geography":["KR"],"entities":[{"name":"Hong Young-jun","canonical_id":"person:kr:hong-young-jun","type":"person","role_in_article":"primary_subject","metadata":{"parent":null}},{"name":"Park Jin-hwi","canonical_id":"person:kr:pak-jin-hwi","type":"person","role_in_article":"quoted","metadata":{"parent":null}},{"name":"Baek Ji-hye","canonical_id":"person:kr:baek-ji-hye","type":"person","role_in_article":"quoted","metadata":{"parent":null}},{"name":"Kim Sun-kyu","canonical_id":"person:kr:kim-sun-gyu","type":"person","role_in_article":"quoted","metadata":{"parent":null}},{"name":"Gwedo","canonical_id":"person:kr:gwido","type":"person","role_in_article":"quoted","metadata":{"parent":null}},{"name":"Seoul National University","canonical_id":"org:kr:seoul-national-university","type":"organization","role_in_article":"mentioned","metadata":{"parent":null}},{"name":"Chung-Ang University","canonical_id":"org:kr:chung-ang-university","type":"organization","role_in_article":"mentioned","metadata":{"parent":null}},{"name":"Korea Astronomy and Space Science Institute","canonical_id":"org:kr:korea-institute-of-astrophysics-and-geophysics","type":"organization","role_in_article":"mentioned","metadata":{"parent":null}},{"name":"Eigen Science","canonical_id":"corp:kr:eisen-science","type":"company","role_in_article":"mentioned","metadata":{"parent":null}},{"name":"DGIST","canonical_id":"org:kr:daegu-gyeongbuk-institute-of-science-and-technology","type":"organization","role_in_article":"mentioned","metadata":{"parent":null}},{"name":"2026 Korea AI Innovation Talent Connect","canonical_id":"org:kr:2026-napatg-aipubssa-cokok","type":"organization","role_in_article":"source","metadata":{"parent":null}}],"claims":[{"id":"c1","statement":"On the 31st of last month, a discussion titled 'How far can AI change the role of scientists?' was held at the '2026 Korea AI Innovation Talent Connect'","as_of":"2026-07-31","as_of_explicit":false,"as_of_raw":"the 31st of last month","source_type":"journalist_analysis","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c2","statement":"AI performs tasks such as mathematical proofs and drafting papers, finding rare phenomena in astronomical observation footage, and generating large quantities of candidate drug compound combinations","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"journalist_analysis","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c3","statement":"As the amount of results generated by AI increases, the number of things that need to be evaluated for correctness and usefulness has also surged","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"journalist_analysis","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c4","statement":"Even if tens of thousands of candidate drug compounds can be generated, deciding which one or two to actually put into experiments remains difficult","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"journalist_analysis","comparison":null,"type":"fact","figures":{"value":10000,"unit":"count","approximate":true,"converted":null},"expiry_hint":null,"insight":null},{"id":"c5","statement":"Rather than eliminating scientists' work, AI is automating the intermediate stages of the research process, and the scientist's role is shifting toward setting questions and hypotheses at the front end of research, and evaluation, verification, interpretation, and persuasion at the back end","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"journalist_analysis","comparison":null,"type":"opinion","figures":null,"expiry_hint":null,"insight":{"analyst_name":"토론 패널들","analyst_affiliation":"2026 대한민국 AI혁신인재 커넥트","confidence":"high","forecast_horizon":"long_term","sentiment":"neutral","reasoning":"AI is merely an automation tool; humans must still handle the core tasks of setting questions and verifying results"}},{"id":"c6","statement":"With the application of generative AI, the cost of creating and generating new things has become extremely cheap, and the bottleneck has shifted toward evaluating, selecting, and verifying","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"unnamed_source","comparison":null,"type":"opinion","figures":null,"expiry_hint":null,"insight":{"analyst_name":"김선규","analyst_affiliation":"아이젠사이언스","confidence":"high","forecast_horizon":"short_term","sentiment":"neutral","reasoning":"The reduction in generation costs has made the evaluation and verification stage a new bottleneck"}},{"id":"c7","statement":"While data-driven methods may dominate in ultra-short-term rainfall forecasting, mid-to-long-term global forecasts over periods of six months or a year require models, as data alone is insufficient","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"unnamed_source","comparison":null,"type":"fact","figures":{"value":6,"unit":"month","approximate":false,"converted":null},"expiry_hint":null,"insight":{"analyst_name":"홍영준","analyst_affiliation":"서울대학교","confidence":"high","forecast_horizon":"long_term","sentiment":"neutral","reasoning":"For problems involving the accumulation of physical phenomena over long periods, existing computational and analytical methods (models) remain valid"}},{"id":"c8","statement":"If a doctor indicates that, out of a one-minute EEG recording, the middle 30 seconds should be examined, using only that 30-second segment as data can significantly boost actual performance","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"unnamed_source","comparison":null,"type":"fact","figures":{"value":30,"unit":"second","approximate":false,"converted":null},"expiry_hint":null,"insight":{"analyst_name":"박진휘","analyst_affiliation":"중앙대학교","confidence":"medium","forecast_horizon":"short_term","sentiment":"bullish","reasoning":"Adding domain knowledge (hypotheses) to data creates more effective models"}},{"id":"c9","statement":"Running an AI model for about an hour to generate substances can produce 10,000 or even tens of thousands of results","as_of":"2026","as_of_explicit":false,"as_of_raw":"about an hour","source_type":"unnamed_source","comparison":null,"type":"fact","figures":{"value":10000,"unit":"count","approximate":true,"converted":null},"expiry_hint":null,"insight":null},{"id":"c10","statement":"Now that AI is entering industry, the ability to persuade is essential; what matters is not just how well AI explains itself, but how the results are ultimately presented and utilized","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"unnamed_source","comparison":null,"type":"opinion","figures":null,"expiry_hint":null,"insight":{"analyst_name":"김선규","analyst_affiliation":"아이젠사이언스","confidence":"high","forecast_horizon":"short_term","sentiment":"neutral","reasoning":"In industrial settings, expert interpretation and persuasion by researchers become more important than AI's automatic explanations"}}],"ai_emotional_context":{"valence":0.2,"arousal":0.3,"primary_emotions":[{"emotion":"calm","intensity":0.7},{"emotion":"optimistic","intensity":0.4}],"secondary_emotions":[{"emotion":"skeptical","intensity":0.2}],"emotional_triggers":[]},"image":{"url":"https://onjblseywainslkhvkav.supabase.co/storage/v1/object/public/article-images/f5e5067b-566e-4dab-a406-318b4e565c8b.jpg","alt":"2026 대한민국 AI혁신인재 커넥트 행사에서 패널 5명이 무대에 앉아 토론하는 모습.","caption":"지난달 31일 ‘2026 대한민국 AI혁신인재 커넥트’에서 열린 ‘AI는 과학자의 역할을 어디까지 바꿀 수 있을까?’ 토론은 AI를 활용하기 시작하면서 바뀐 연구 방식과 이후 과학자에게 남는 역할을 짚었다. (사진=테크42)","source":"first_img","alt_status":"auto"}},"provenance":{"source_chain":["primary_reporting"],"original_source_url":null,"related_articles":[]},"temporal":{"freshness":"recent","next_update_expected":null},"access":{"license":"neupai_standard","attribution_required":true,"structured_data":"free","full_text_available":false,"full_text_access":null}}