This page discusses artificial intelligence in higher education. The focus of the content is on the use of artificial intelligence in information search, but we also highlight the ethical issues raised by artificial intelligence. Artificial intelligence and its applications are developing rapidly, in the same way this guide is constantly under development, check for updates and follow the development of the field.
Artificial intelligence is a broad concept that refers to the ability of computer systems to perform tasks and processes that normally require human intelligence. Such tasks can include, for example, image recognition, speech recognition, problem solving and decision-making. Artificial intelligence is already used in many ways in our everyday lives: in recommendation lists provided by various applications, in language translation, in targeting advertising and predicting housing prices, etc.
Traditional information search in databases is based on search terms and logical operators (e.g. AND, OR, NOT). The searcher controls the process: he or she defines the search terms, delimitations and selects the sources themselves. The results are documents or references based on pre-indexed content, and the user can check the original source. You can read more about the subject from library's Information searching guide.
Generative AI cannot really be called information searching. AI does not search for information directly from sources, but generates answers based on its training data and the model's internal structures. The AI's language model forms the answers by generating text, i.e. predicting the next words based on probabilities using the data it has been taught. Language models are so good here that the answer can sound completely plausible, even though it is not based on anything. In this case, we are talking about hallucination, i.e. incorrect or misleading information produced by AI. "Information searchl" here is more of a conversation than a search: the user gives a prompt, and the model interprets the prompt and responds to it according to the context. Purely generative AI services include Claude, ChatGPT, and Gemini without using web search.
If an AI service uses retrieval-augmented generation (RAG), it combines traditional information search and generative AI. First, the service searches for information from external sources, such as websites, databases or documents, using traditional search. The service then provides the best search results to a language model, which then generates an answer based on them. This way, the answers are based on information retrieved from elsewhere, which can make them more up-to-date, more accurate and the sources are usually known. However, it is worth noting that a service using retrieval-augmented generation can also hallucinate. It can use the information it has extracted from the source incorrectly and formulate an answer that no longer corresponds to the original content or context. Services that use retrieval-augmented generation from the internet include Copilot, Perplixity AI and ChatGPT, as well as Gemini, when they have the search feature enabled. Scopus AI, on the other hand, searches information from the curated scientific reference database Scopus.
