Artificial intelligence can be a tricky concept to grasp, with scientists often using complex jargon to explain their work. This leads to confusion for many of us who are not well-versed in the field. To help you navigate this convoluted world, we’ve put together a glossary of key terms used in the artificial intelligence industry. We’ll be updating this glossary regularly to include new terms as researchers make groundbreaking discoveries and uncover potential safety risks. Let’s dive into some of the important terms you might come across in articles about AI.
AGI, or Artificial General Intelligence, is a term that refers to AI systems that are more capable than the average human at various tasks. Different experts have slightly different definitions of AGI, but they all point to AI that can outperform humans in many economically valuable tasks. AI agents are tools that use AI technologies to perform tasks on your behalf, going beyond basic chatbots to handle more complex activities like booking tickets or writing code. These agents rely on multiple AI systems to carry out multistep tasks, but the infrastructure is still being developed to fully realize their potential.
Deep learning is a subset of machine learning that uses artificial neural networks to make complex correlations in data. These algorithms can identify important features in data without human input, learning from errors to improve their performance over time. Diffusion and distillation are techniques used to generate realistic data and extract knowledge from AI models, respectively. Fine-tuning involves further training an AI model for specific tasks, while GANs use a competitive framework to generate realistic outputs. Hallucinations in AI refer to models making incorrect information, leading to potential risks. Inference is the process of running an AI model, and transfer learning allows knowledge gained from one model to be applied to a new task. Weights are numerical parameters that determine the importance of features in training AI models, shaping their outputs.
