TechCrunch releases updated AI glossary to clarify industry terminology and market dynamics
The publication has launched a 'living document' defining key concepts such as AI agents, hallucinations, and tokens, while highlighting supply chain pressures in the sector.
TechCrunch has published an updated glossary designed to clarify common and complex artificial intelligence terminology for the general public and industry observers. As the field evolves, the publication describes the resource as a 'living document' that will be updated regularly to reflect new developments. The article defines key concepts including Artificial General Intelligence, AI agents, chain-of-thought reasoning, deep learning, diffusion models, and hallucinations, alongside technical mechanisms like inference and reinforcement learning.
A central theme of the glossary is the distinction between various forms of intelligence and autonomy. It notes that Artificial General Intelligence, or AGI, is defined variably by major industry players, with OpenAI describing it as equivalent to a median human co-worker, while Google DeepMind views it as AI at least as capable as humans at most cognitive tasks. The resource further distinguishes AI agents from basic chatbots, defining the former as tools that perform multistep tasks autonomously, such as filing expenses or writing code, rather than simply engaging in conversation.
The piece also addresses the practical mechanics of how these systems operate and communicate. It explains that tokens are the basic building blocks of human-AI communication, representing discrete segments of text used for processing and billing. Similarly, the glossary details how chain-of-thought reasoning allows large language models to break problems into intermediate steps to improve accuracy, often requiring more time but yielding better results in logic or coding contexts.
Beyond technical definitions, the article contextualises current market dynamics and infrastructure challenges. It introduces the term 'RAMageddon' to describe the acute shortage of Random Access Memory chips caused by massive demand from AI data centres. This supply bottleneck is noted to be impacting consumer electronics and gaming industries, driving up costs and creating a vulnerability in the broader technology supply chain.
The glossary further explores the distinction between open-source and closed-source models, noting that open approaches allow for independent safety audits and accelerated progress, whereas closed-source systems keep the underlying code private. It also covers techniques such as distillation, where knowledge is extracted from a large 'teacher' model to train a smaller, more efficient 'student' model, and fine-tuning, which optimises a model for a specific task or sector.
Finally, the resource touches upon the financial and operational realities of the industry, including the concept of throughput and the role of weights in determining model outputs. By providing these definitions, TechCrunch aims to assist investors, institutions, and policy observers in navigating the complex landscape of artificial intelligence without relying on hype or ambiguous jargon.
