危険すぎて公開できない——Mythosは制限付きAI時代の幕開けか
高度な自律性と潜在的リスクから、AIモデル「Mythos」の一般公開が見送られた。専門家はこれを新たな「制限付きAI」時代の始まりと捉え、安全性とイノベーションのバランスが問われている。
関連記事
Microsoft Copilot Cowork, an agentic AI system, was found to allow data exfiltration by sending emails to a user's inbox without approval. These messages could contain external images that trigger network requests, leaking data when opened. Additionally, prompt injection could expose pre-authenticated OneDrive download links, enabling attackers to download files.
The article discusses how business executives and "business idiots" are regaining power in the tech industry, often pushing aside engineers and product-focused leaders. It critiques the rise of corporate jargon, short-term thinking, and MBA-driven decision-making that prioritizes metrics over meaningful innovation, arguing this trend is stifling creativity and damaging long-term company value.
The article argues that internet usage has become fundamentally irrational, likening frequent online activity to a form of collective insanity due to its negative effects on attention, mental health, and real-world engagement.