Artificial intelligence is systems that can perform tasks that normally require human intelligence. It can be understanding text, recognizing patterns, proposing solutions or generating content based on data.
An algorithm is a step-by-step recipe that describes how to solve a problem. In AI, algorithms are used to analyze data and make calculations that lead to a result.
Machine learning is a branch of AI where systems learn from data instead of being programmed with fixed rules. The model gets better the more relevant examples it is trained with.
Deep learning is a type of machine learning that uses neural networks with many layers. The method is particularly effective for complex tasks such as language comprehension, image recognition and speech recognition.
A language model is an AI model trained to understand and generate text. It works by predicting the next word or token in a text based on context and probabilities.
Generative AI is artificial intelligence that can create new content such as text, images, sound or code. The output is not copies, but statistically generated content based on patterns in training data.
Computer vision is AI that can analyze and interpret images and video. The technology is used to identify objects, read text in images or recognize visual patterns.
Speech-to-text is technology that automatically converts speech into written text. It is used, among other things, for transcription of meetings, dictation and accessibility solutions.
Text-to-speech turns written text into artificial speech. It is often used in voicebots, reading content and solutions for users with reading difficulties.
A chatbot is a conversation-based interface where the user communicates with a system via text or speech. Chatbots can answer questions, guide users or perform simple tasks.
An AI agent is a more advanced form of AI that can work towards an outcome and perform multiple steps. It can combine responses, actions and tools into one coherent process.
A prompt is the instruction or query that the user gives to an AI. The wording of the prompt is of great importance for how and how precisely the AI responds.
Prompt engineering is the work of systematically formulating prompts so that the AI delivers more precise, useful and consistent answers. It can include structure, examples and clear requirements.
Input is the information that is given to the AI, e.g. text, data or files. Output is the result that the AI returns, e.g. an answer, a summary or a suggestion.
The user interface is where the user interacts with the AI system. It can be a chat, an app or an embedded solution on a website.
Datasets are the data on which an AI model is trained. The quality, relevance and composition of data are of great importance for the model's results and limitations.
A model is the fully trained AI system that can take input and produce output. The model contains the patterns and relationships learned from training data.
Inference is the process by which a trained AI model is used to provide answers to new inputs. It is the actual "time of use" when the model produces output.
Bias are systematic biases in AI systems that can lead to unreasonable or misleading results. Bias often occurs due to skewed or incomplete training data.
Hallucination occurs when the AI generates answers that sound convincing but are factually incorrect or made up. This is a known limitation of generative models and cannot be avoided since the AI in principle does not know if it is telling the truth.
Black box is a general term, but is currently mainly used to describe AI systems where it is difficult to explain how a certain output has been reached.
Human-in-the-loop means that a human is actively involved in the process and has the opportunity to check, approve or reject the AI's output before it is used.
