Columns– category –
-
Altman lays out three paths to AI safety, calls for speed limits on self-improvement
OpenAI CEO Sam Altman reportedly laid out three possible end states for AI safety during an internal talk. This article examines the differences among the three proposals—an international oversight body, a merger of leading labs, and one lab pulling ahead—through the lens of governance choices. -
AI assistant memory test: retrieval outperforms full-history approach, with GPT-4o mini achieving an 88.9% success rate
Does giving an AI assistant its entire lengthy conversation history make it smarter? In a small-scale test using a fictional person, an approach that searches for and supplies only the relevant parts achieved both a higher success rate and lower cost than one that provided the full history. -
Passing on the Specialized Skills Behind Arita Ware to the Next Generation: AI Transcription as a Recordkeeping Aid
Arita ware, made in and around Arita Town in Saga Prefecture, is porcelain produced through a division of labor. This article considers how AI transcription and other tools can document craftspeople’s judgments and knowledge of each stage, and pass them on to the next generation—not by replacing people, but by supporting them.
1