Jargon:Generative AI
1 min read
Say this at the party
“It's AI that makes new things (writing, pictures, music, code) instead of just sorting or checking things that already exist.”
In plain English
Most older AI was a judge. It looked at something and gave a verdict: spam or not spam, cat or dog.
Generative AI is a maker. It has studied huge piles of examples, learned the patterns, and uses them to produce something new when you ask: an email, a drawing, a song, a bit of code.
It doesn't copy and paste. It builds each answer fresh, which is why it can be brilliant one minute and confidently wrong the next.
Where it came from
Computers have generated text and art since at least the 1970s. The modern version took shape in 2014, when new methods such as GANs let neural networks make convincing images. The 2017 transformer design then led to the GPT models, and the term went mainstream with the AI boom of the early 2020s.
See it in action
Priya types "write a polite note to my neighbour about their hedge" into ChatGPT. Ten seconds later she has three versions to choose from, none of which existed before she asked.
What people get wrong
- It isn't looking answers up like a search engine. It's predicting what a good answer would look like, so it can make things up.
- Generative AI is one kind of AI, not all of it. Your bank's fraud checker is AI too, but it doesn't generate anything.
Go deeper
- ArticleWikipediaGenerative AIA full overview, with a clear history from early text generators to today's chatbots and image makers.
- PaperGoodfellow et al., 2014Generative Adversarial NetworksThe 2014 paper that taught neural networks to make realistic images by pitting two of them against each other.
- GuideIBMWhat is Generative AI?A friendly business-level explainer of how generative AI works and where it is used.
Related jargon
Drafted with AI. The AI Council (Claude, ChatGPT and Gemini) gave counsel. Phin signed it off.