AI Glossary 2026: AGI, Agents, and Chain of Thought Explained

The field of artificial intelligence is evolving at breakneck speed, with new specialized terms emerging daily. From acronyms such as LLM (Large Language Model), RAG (Retrieval-Augmented Generation), and RLHF (Reinforcement Learning from Human Feedback) to concepts like AGI (Artificial General Intelligence) and AI agents, these terms often dominate product meetings and investor presentations. Even tech professionals sometimes find them overwhelming.

On July 3, 2026, TechCrunch addressed this challenge by publishing a glossary that provides clear, concise definitions of key AI terms likely to be encountered in the field. This article builds on their work, presenting definitions and context for these terms in Japanese. The glossary is positioned as a “living document” that will be regularly updated to reflect the rapid evolution of AI systems.

The Definition and Ambiguity of AGI

AGI (Artificial General Intelligence) is one of the most hotly debated terms in the AI field. According to the TechCrunch article, AGI refers to “an AI that is better than the average human at many or most tasks.” However, its definition varies slightly across organizations.

Sam Altman, CEO of OpenAI, once described AGI as “an entity equivalent to an average human that you could hire as a colleague.” Meanwhile, OpenAI’s charter defines AGI as “highly autonomous systems that outperform humans at most economically valuable work.” Google DeepMind’s interpretation differs slightly, considering AGI as AI with capabilities at least equivalent to humans in most cognitive tasks.

TechCrunch notes that even experts at the forefront of AI research are confused by these varying definitions, suggesting that the lack of a standardized definition highlights both the immaturity and potential of this field.

The Essence of AI Agents

AI agents go beyond the capabilities of basic chatbots, serving as tools that execute a series of tasks on behalf of users. These tasks could include submitting expense reports, booking tickets or restaurant reservations, and even coding or maintenance tasks, often involving multiple steps.

However, this area is still in its developmental stage, with infrastructure yet to be fully established. According to TechCrunch, the term “AI agent” might mean different things to different people. At its core, it refers to autonomous systems that leverage multiple AI technologies to complete multi-step tasks.

Investment in autonomous systems is also accelerating in robotics, as evidenced by Hyundai’s acquisition of Boston Dynamics as a wholly owned subsidiary (Hyundai Fully Acquires Boston Dynamics; SoftBank Withdraws for $325 Million). However, current AI agent implementations are predominantly software-based.

API Endpoints and Automation

TechCrunch describes API endpoints as “buttons” behind the scenes of software. Other programs can “press” these buttons to perform tasks like authentication, data retrieval, or executing functions. Developers use these interfaces to build integration between applications.

Many smart home devices and connected platforms have these hidden buttons that general users are unaware of. With advancements in AI agents, these agents can autonomously discover and use API endpoints without human intervention. While this opens up powerful automation possibilities, it also brings risks of unexpected behavior.

The Mechanism of Chain of Thought

Chain of Thought mimics the human process of solving complex problems by writing down intermediate steps. For straightforward questions like “Which is taller, a giraffe or a cat?” humans can answer without much thought. However, for problems requiring mental calculation or logical reasoning, explicitly writing out intermediate calculations or steps improves accuracy.

It is known that prompting AI models to output such sequential reasoning significantly enhances accuracy on complex problems. While the TechCrunch article mentions specific implementations and applications of this method, the provided text leaves out some details. Generally, Chain of Thought is widely used as a prompting technique to enhance the reasoning capabilities of large language models, with proven effectiveness in benchmarks like math problems and common-sense reasoning.

The Social Significance of the Glossary

The flood of AI-related terminology reflects the rapid pace of technological advancement, far outstripping traditional software development. Without standardized definitions from regulatory bodies, vendors and research institutions propagate their interpretations, leading to confusion as the same terms acquire different meanings depending on the context.

The TechCrunch glossary serves as a response to this confusion. The article explicitly states that the glossary will be continuously updated, suggesting that the AI field requires dynamic, adaptable documentation rather than static definitions to keep pace with its evolution.

Editorial Opinion

In the short term, the lack of unified definitions for AI terms among companies undermines market transparency. The ambiguity surrounding AGI, in particular, complicates decision-making for investors and regulators. The coming three to six months will be critical in determining whether the industry accelerates efforts toward a common understanding or whether standardization bodies will need to intervene. Similarly, the varying definitions of AI agents contribute to product differentiation but can also lead to confusion.

In the long term, progress in standardizing terminology could enable better product comparisons and regulatory frameworks. However, as technological evolution continually outpaces definitions, maintaining a “living document” will require ongoing expert oversight and updates. In this context, the role of media outlets like TechCrunch is significant and commendable.

The editorial team believes that further discussion is needed on how well readers understand these terms and on real-world instances of confusion. Particularly, differing perceptions of the scope of AI agent autonomy directly tie into practical issues like security and accountability, making it a topic that warrants closer examination.

References

• TechCrunch AI — Published on 2026-07-03