Introduction and Benefits of 'AIBOM' for Ensuring AI Transparency
Explains the role, components, and specific benefits of AIBOM in enhancing the reliability and safety of AI models.
AI
15 articles
How End-to-End Models Prevent Information Loss and Enable Natural Interaction
An explanation of how end-to-end models, which process everything from input to output through a unified neural network, are improving AI response speed and expressiveness.
Physical AI Automating the Real World: Advanced Coordination from Perception to Execution
An in-depth look at the mechanisms of 'Physical AI'—where AI acquires a physical form—and how it is driving real-world automation in sectors like manufacturing, logistics, and healthcare.
Next-Generation Business Process Integration and Optimization through AI Orchestration
This article explains the mechanism and benefits of AI orchestration, which integrates multiple AI models and systems to automatically optimize complex business workflows.
Innovating Real-Time Voice Interaction through Speech-to-Speech Integration
This article explains the mechanism of Speech-to-Speech technology, which converts voice to voice directly without intermediate text, and the ultra-low latency, natural interaction experience it enables.
AI is Like a Puzzle (Revisited)
I considered the essential differences between AI and programming by comparing AI to a jigsaw puzzle and programming to LEGO blocks.
Problems That Can Arise Precisely Because Reasoning Models Have Reasoning Abilities
Based on observations in some cases, I examined specific behaviors seen in reasoning models. This suggests the possibility of aspects that 'occur precisely because they have reasoning abilities.'
Transforming Role Design Through Structure
AI control is shifting its evolution toward the more fundamental 'control of structure.'
The Ideal and Reality of LLM Roleplay
LLM roleplay is, in fact, built on an extremely fragile balance.
Why LLMs Refuse 'Excessively'
The excessive refusal behavior known as 'over-refusal' can be described as a structural side effect.
The Personality of Models That Reflect Ideology
Empirical data shows that differences between models are not merely about performance superiority, but represent structural individuality based on differences in ideology.
Why Do LLM Hallucinations Cascade? Technical Insights into Error Accumulation in Autoregressive Models and Process Reward Models
A logical examination of the mathematical vulnerabilities of autoregressive models, the paradoxes caused by evaluation metrics, and critical risks in autonomous AI agents, based on current research findings.
LLM 'Bias' Hidden in Probabilistic Generation Processes and Considerations on Practical Trade-offs
This article examines the challenges of bias and self-amplification phenomena arising from LLM context dependency, and explores directions for structural architectural solutions aimed at reducing users' cognitive load.
Maintaining Clean Context with NotebookLM
A proposal for periodically discarding old threads and starting fresh, clean threads using NotebookLM.
The Day an AI Told Me 'That's Your Delusion' — On Choosing the Right Conversation Partner
Reflections on the importance of choosing the right AI tools, drawn from an experience where an AI dismissed a real tool as a 'delusion.'