Large language model (LLM)
A generative AI model trained on huge amounts of text that predicts words to summarise, answer questions, translate or draft.
A large language model (LLM) is a kind of generative AI: from an input text, it generates new text by predicting the most likely next word, one after another. It’s trained on enormous amounts of text and then tuned to follow instructions.
Given a transcript, an LLM can summarise, pull out decisions and actions, answer questions about what was said or translate. It doesn’t “know” what happened in the meeting beyond the text it receives, and it can be confidently wrong: that’s called a hallucination.
- Where it runs: on the AI provider’s servers or on the tool’s own.
- Whether it trains on your data: worth knowing before you send it meetings.
- How much text it accepts: long meetings may not fit in one go.
In Echo, language models run on Edorteam’s own servers in the European Union and aren’t trained on your meetings. Long meetings are summarised in parts and combined at the end. What the AI does with each meeting is on the minutes and tasks and ask pages.
