South Korea's Electronics and Telecommunications Research Institute (ETRI) said Tuesday it will pursue a "director-level AI scientist" capable of autonomously forming hypotheses, designing experiments, analyzing data, and verifying results by 2035. The announcement came at ETRI Conference 2026 in Seoul, where the institute outlined a new R&D strategy built around AI and digital transformation.
The goal marks a shift from AI as a research assistant to AI as a research director that manages the full scientific cycle without human intervention. Park Se-ung, president of ETRI, said the institute will "combine the ICT capabilities accumulated over the past 50 years with AI to create new growth engines for South Korea."
A three-stage path to autonomous research
ETRI's plan, which it calls an "AI scientist for all," unfolds in three phases. The first is an AI research companion, where human researchers and AI jointly form hypotheses and run experiments. The second is a human-supervised autonomous science system, in which AI operates scientific agents and laboratories based on direction from researchers.
The final phase is a fully autonomous, full-cycle research system where AI manages multiple research projects simultaneously. ETRI said it will build a Korean AI scientist platform combining a science-specialized foundation model, expert AI agents, hypothesis generation tools, experiment design capabilities, a scientific AI operating system, and autonomous laboratory functions.
The institute also said it aims to pursue Nobel Prize-caliber research outcomes in connection with the national "K-Moonshot" initiative.
AI data centers as 'token factories'
ETRI is also targeting core technology for AI data centers (AIDC), which it describes as "token factories" that convert power and data into AI computing capacity. Park framed the AIDC as "the heart of the AI era," drawing a parallel to power plants in the industrial age.
The institute plans to integrate AI semiconductor, optical interconnect, optical network, and AI computing software technologies. Key milestones by 2031 include a 100 Tbps optical I/O chiplet and a 100 Tbps optical transceiver. From 2032, ETRI plans to expand toward a "collaborative AIDC" linking ground-based computing resources with those in space.
For researchers following developments in AI for Science & Research, the ETRI roadmap offers a concrete timeline for how autonomous research systems are being prioritized at the national level.
National mission R&D framework
The conference marked ETRI's 50th anniversary and introduced a new vision: "A warm world where we live together, led by AI and DX innovation." The institute said it will concentrate research capabilities in five areas: AI-native 6G and satellite communications, safe artificial general intelligence, immersive spatial media, physical AI, and public and industrial AI and DX.
Park said ETRI will shift to a national mission-centered research framework, consolidating R&D around AI and digital transformation. He added that the institute would expand cooperation with industry, academia, research institutions, and global partners. The goal, he said, is for South Korea to emerge as one of the world's top three AI powers.
Why this matters for science and research professionals
ETRI's 2035 target for a director-level AI scientist signals that research institutions are now planning for AI systems that do not just accelerate individual tasks but own entire research workflows. For working scientists, the practical question is how roles shift as hypothesis generation and experiment design become partially or fully automated. Building skills in directing AI research systems - rather than simply using them for analysis - is likely to become a distinct competency. The AI Learning Path for Research Scientists addresses this transition directly, covering the skills needed to supervise and collaborate with increasingly autonomous research tools.
Your membership also unlocks:
