Generative AI is artificial intelligence that creates new content, such as text, images, audio, video or computer code, in response to instructions called prompts. Popular chat assistants and image generators are examples. Instead of only analysing data, generative AI produces something new.
What generative AI is
Traditional AI often classifies or predicts: is this email spam, will this customer leave? Generative AI goes further and produces original output: a summary, an email draft, an illustration, a piece of music. It learns the patterns of language, images or sounds from enormous amounts of training data, then uses those patterns to generate new content that fits a request.
How it works
Most text-based generative AI is built on large language models (LLMs), which use an architecture called the transformer, introduced by researchers in 2017. During training, the model reads vast amounts of text and learns to predict the next word in a sequence. Repeated billions of times, this simple task produces a model that can write, summarise, translate, answer questions and write code.
Image generators often use diffusion models, which learn to turn random noise step by step into an image matching a description. Many newer systems are multimodal, working with text, images and audio together.
A brief history
Generative AI existed in research for years, but it reached the mainstream when easy-to-use chat assistants were released to the public in late 2022. Since then, the technology has spread rapidly into search engines, office software, coding tools, customer service and creative work.
Everyday uses
People use generative AI to draft emails and reports, summarise long documents, brainstorm ideas, translate languages, write and fix code, create images for presentations and learn new topics. Businesses use it for customer support, marketing content and analysing documents.
Limitations
Generative AI can produce confident but wrong answers, often called hallucinations. It may reflect biases in its training data. Questions about copyright, privacy and the use of training data are being debated and tested in courts in several countries. Its output should be reviewed, especially for facts, numbers and important decisions.
Why it matters for investors
Generative AI is driving large investments in computing infrastructure, such as data centers, chips and power, and is changing how many companies operate. For investors, understanding it helps in following a major economic theme, while staying alert to hype.
Using generative AI well
Clear prompts produce better results: say what you want, give context, specify the format and length, and provide examples where helpful. Ask the AI to show its reasoning or list its sources, then check them. Treat the first answer as a draft to improve, not a final product. And avoid entering confidential personal or business information into tools whose data policies you do not know.
Generative AI and business
Companies are adopting generative AI for customer support, document processing, marketing and software development. The biggest gains so far often come from speeding up routine work rather than replacing whole jobs.
Common misconceptions
"Generative AI understands like a human." It generates text based on learned patterns, not human understanding.
"If it sounds confident, it's correct." Confidence and accuracy are not the same.
"It copies existing content." It usually generates new combinations, though it can sometimes reproduce training material.
Frequently asked questions
What is a prompt?
The instruction or question you give to a generative AI system.
What is an LLM?
A large language model, the technology behind most AI chat assistants.
What is a hallucination?
A confident but incorrect or invented answer from an AI system.
Can generative AI give investment advice?
It can explain concepts, but its output should not replace research or professional advice.
How PositoAI fits in
PositoAI uses AI to write short company summaries from recent news and regulatory information, so you can understand a company's situation quickly. Read more in our guide to AI.
Summary
Generative AI creates new content, such as text, images, audio and code, from prompts. Text systems use large language models based on transformers; image systems often use diffusion models. It is powerful and widely used since 2022, but it can hallucinate and reflect bias, so its output should always be checked.