Intro
Connecting the dots!
Research
Conferences
I attended the International Conference on Next Horizons in Polymers (IUPAC-PSK50) at BEXCO in Busan from September 29 to October 1, 2026. Marking the 50th anniversary of the Polymer Society of Korea, the conference featured a broad program covering contemporary developments in polymer science and engineering, including Polymers for Electronics and Photonics and AI-assisted Design and Simulation of Polymers.
The event provided an opportunity to explore current technologies and industry trends through exhibitions, workshops, and presentations. Brief booth visits included Alchemist Inc., Aifrica, Apple Manufacturing R&D Accelerator, HP Korea, FainRobotics, and PUDU Robotics, while more detailed technical exchanges took place with AiM Future, CAST AI, ENERZAi, RealMan Robotics, Seagate, and You.com.
At GASKAIC 2026 in Daejeon, I spoke about a concern I have with today’s AI narrative: the greatest danger may not be that AI will make humanity unnecessary, but that people will be persuaded to believe it will. AI models can already outperform humans in many defined tasks, yet that is very different from independently recognizing new societal problems, deciding what matters, and taking responsibility for solving them.
Blog
Notes and perspectives.
The news that OpenAI Prism, the advanced LaTeX workspace, will shut down on January 1, 2027, has raised serious questions about the fate of researchers' unpublished scientific work and the transparency of the platform's data practices. But for those of us who actually used it, the sudden announcement of its planned closure feels like the final piece of a calculated playbook.
ኣብይ ስልጣን ክሕዝ ናይ liberal political ideological and/or free market economy ሽፍን ተጠቂሙ 'ዩ። ኮይኑ ግን ናይ ባዕሉ ኣጀንዳ ነይረምዎ እዩም! መሰረቶም 'ወን ታርካዊ ፣ ወልቀ ባህግታትን ቂም-በቀልን 'ዮም።
The development of AI technology should not stop; instead, it should be properly regulated, and there should be greater transparency about the data used to train these systems. If an AI is trained on information about how to build bombs, develop nuclear or biological weapons, or carry out cyberattacks, it may become capable of assisting with those activities.
The 2026 DGIST AI Build Week
AI models are not built from nowhere; they are trained on the accumulated scientific, engineering, mathematical, software, and educational commons produced by people around the world for centuries. When global knowledge, open-source code, public research, and collective human labor become the input, democratic access must be part of the output.