10 terms that everyone needs to know in the age of AI

By Vijay Anand

CNBC-TV18.com

Published July 11, 2024

Artificial intelligence (AI): A super-smart computer system that can imitate human tasks like understanding language, decision-making, translation, sentiment analysis, and learning from experience. AI processes large datasets through algorithms to create models that automate tasks requiring human intelligence. 

Machine Learning (ML): An AI technique where computer systems learn to identify patterns and make predictions by running data through algorithms repeatedly, enabling complex problem-solving.

Large Language Models (LLMs): Large language models use machine learning and neural networks to mimic human communication. Trained on vast text data, they can translate, answer questions, summarise, and generate content.

Generative AI (Gen AI): Generative AI uses large language models to create new content, not just provide information. It learns patterns to generate novel pictures, music, text, videos, and code.

Hallucinations: Generative AI can create content, but may produce inaccurate "hallucinations" due to lack of grounding in real information. Developers address this by providing additional trusted data.

Responsible AI: Responsible AI ensures safety and fairness at all levels — model, software, interface, access. Key is understanding training data to mitigate biases and better represent society.

Multimodal models: Multimodal models can work with diverse data like images, sounds, and text, combining this information to perform complex tasks like answering questions about images.

Prompts: Prompts are instructions that tell AI models what task to perform. Carefully designing prompts is crucial, like specifying details when ordering, to achieve the desired outcome.

Copilots: A Copilot is a digital assistant powered by large language models to help with tasks like writing, coding, and analysis, while maintaining safety and user control.

Plugins: AI plugins act like apps, enabling specific functionalities without altering the core model. They integrate AI deeper into the digital ecosystem, enhancing its capabilities.