Artificial Intelligence (AI) models human neural network activity and aims to mimic human consciousness and the way humans perform tasks.
Language models are AI applications that use deep learning methods and neural network structures to learn language structures and rules, and to predict and produce linguistic content in different contexts. A language model does not contain its training data and does not "know" anything directly. It is a statistical model built from data that tries to predict the next word from a given text.
Language models are only part of the AI world and many different everyday services such as search engines, online shops, map services and language translators nowadays have AI algorithms of varying levels.
AI is used by giving the application you are using a prompt.
The precise wording and clarity of prompts are key to ensuring that the AI understands the user's intent and produces the desired response or content. Prompts can be modified and can be used as a kind of conversation with the AI application.
Here are some tips on how to create functional and personalised content using prompts.
Remember that an AI application may sometimes need more precise guidance or more attempts to produce the best possible response. Prompts provide the guidelines for the AI in terms of topics, objectives and priorities.
Even if AI seems to be a fast, efficient and easy way to proceed in scientific information retrieval, it is still advisable to use the university's own licensed scientific databases.
Evaluating the answers produced by AI applications is complicated by a lack of knowledge about what data mass they are based on, how they produce answers, and there is not always a guarantee that the information they provide is up-to-date. However, AI applications can be helpful when you already know a lot about a topic and can critically evaluate their outputs. Different information needs require different information retrieval tools. It is therefore important to first consider what your information needs are and whether AI applications are suitable.
Perhaps the most effective way to combine AI with information retrieval from scientific databases is to ask the AI for a syntax, i.e. an optimised search query for finding information in library databases. In this case, the AI application can add Boolean operators in your search query, and it will also be able to recommend additional keywords for your search.
Always remember that applications also produce incorrect and misleading information, and it is not easy to identify incorrect information if you are unfamiliar with the topic. In this case, the need for source criticism becomes even more important. So check whether the sources provided by AI actually exist and what their contents are.
Perhaps the most effective way to combine AI with information retrieval from scientific databases is to ask the AI for a syntax, i.e. an optimised search pattern for searching information from library databases. Then the AI application can add Boolean operators to your search query and may also recommend additional search terms for your search.
For example, if you want to search for information about Bach's cantatas, you can ask the AI program the question "I want to search for information about Johann Sebastian Bach's cantatas. For the library database, please make me a search query using Boolean operators." The program could, for example, generate a search query such as ("Johann Sebastian Bach" OR "Bach, Johann Sebastian") AND "cantatas" based on this prompt.
For more information on Boolean operators, see the Search techniques page of this guide.
AI generates things, i.e. it simply reflects the input it receives. A constraint on its use is the possible inaccuracy of the data. This means that the AI may "hallucinate" the answer, i.e. produce incorrect information, for example fabricated references.
Information provided by AI may also be damaging or biased due to the source material used to teach the language model. It is worth remembering that most of the materials used to teach language models are of Western origin. This has been found to reduce the global usability of the application.
The responsibility for the accuracy of the written information lies with the AI user. The AI does not know or care whether something is true or not.
Think carefully about what information you enter into the service. AI services are only suitable for processing public data. This means that you should not enter any personal data, business secrets, confidential information or data that should be kept secret, for example.
Any material you enter will be stored and may be used for AI training.
Also, always bear in mind where the information you use comes from. The University of the Arts Helsinki's own rules and ethical guidelines are very important to remember.