Skip to Main Content

AI in information retrieval

Large language models (LLMs)

Much public attention has recently been given to large-scale language models trained by artificial intelligence. Such technology is used by for example ChatGPT. Large language models are not search engines, they produce answers based on the probability of words occurring in sequence. Large language models are trained on large text masses and they generate their answers based on a statistical model. The language model can be interacted with in much the same way as a human, but the answers may also be unreliable or completely made up.


ChatGPT

  • ChatGPT is an OpenAI service and is currently available in free and paid versions
  • The data used to train ChatGPT is not disclosed, the paid version ChatGPT-4 can also use the internet
  • ChatGPT can be used to orientate on the topic, e.g. formulating a research question, searching for keywords and finding suitable tools
  • Note! Unable to provide real sources for the texts they have written. If you ask for a list of sources, the sources are incorrect or completely fictitious.