Artificial intelligence (AI) is a complex and evolving field, often shrouded in technical terminology that researchers and practitioners use to communicate their findings. To assist those navigating this intricate domain, a glossary of key terms associated with AI has been compiled, which will be updated regularly to reflect ongoing advancements and new safety considerations.
Artificial General Intelligence (AGI) refers to a form of AI that matches or exceeds human capabilities across most tasks. This vague term is defined variably by experts; for instance, OpenAI’s CEO Sam Altman views AGI as a median human-like collaborator, while Google DeepMind approaches it as an AI that can competently handle most cognitive tasks.
An AI agent denotes a sophisticated AI tool capable of managing tasks such as booking tickets or writing code, far beyond basic chatbot functions. The exact meaning of "AI agent" can vary, reflecting the ongoing development in this area.
Chain of Thought reasoning represents a critical process for large language models (LLMs) where complex problems are broken down into smaller, manageable steps to produce accurate outcomes, particularly in logic and coding.
Compute refers to the computational power that drives AI operations, encompassing the hardware like GPUs and CPUs necessary for developing and deploying AI models. This computational backbone is essential to the AI industry’s functioning.
Deep Learning is a subset of machine learning that employs multi-layered neural networks to identify complex patterns within vast datasets, allowing AI systems to self-improve and reduce the need for human intervention.
The Diffusion technique is central to generative AIs, simulating a process akin to destruction and reconstruction of data, facilitating the generation of content such as art and music.
Distillation is a method employed to refine a large AI model into a smaller, more efficient one while retaining its essential capabilities, often used to accelerate model development.
Fine-tuning involves further training an AI model on specific tasks using specialised data, enhancing its utility in targeted applications.
A Generative Adversarial Network (GAN) comprises two neural networks—one generating content and the other evaluating it—creating a competitive environment that leads to improved realism in outputs.
Hallucination refers to incidents where AI models produce incorrect or fabricated information, an issue that raises concerns in reliability and safety.
Inference is the execution of an AI model to derive conclusions from previously encountered data, heavily reliant on previous training.
Large Language Models (LLMs) are advanced AI systems powering various applications, characterised by their foundational neural networks that learn from extensive data sources to generate contextually relevant outputs.
Memory Cache technology optimises AI inference by storing previous calculations to improve response times to user queries.
A neural network underpins much of AI’s progress by mimicking the human brain’s interconnected structure, enabling complex data analysis.
RAMageddon describes the current shortage of RAM chips due to burgeoning demand from the AI sector, affecting various industries from gaming to consumer electronics.
Training is the foundational stage of AI development where algorithms learn from data inputs, with costs associated varying based on the complexity of the data required.
Tokens serve as essential data units that facilitate human-AI communication, playing a crucial role in processing queries and informing AI services.
Transfer Learning allows models to leverage previously acquired knowledge for related tasks, improving efficiency, though it often requires additional data for specific applications.
Lastly, Weights are critical parameters that determine the influence of specific features within AI training, shaping the model’s outcomes based on prior learning experiences.
This glossary offers a vital resource for understanding the dynamic landscape of artificial intelligence, which continues to evolve.
Fanpage: TechArena.au
Watch more about AI – Artificial Intelligence
