AI Glossary

Data used continuously during the development of a model to tune its settings and to continuously verify its performance.

An effort to ensure that a model's goals and behavior align with human values and intentions, thereby preventing undesirable or harmful conduct.

A database optimized for storing and searching vector representations (embeddings), which is the basis of vector and semantic search.

Search based on the similarity of semantic representations (embeddings) rather than keyword matching; the basis of modern RAG systems.

The management of risks associated with vendors of AI solutions and services, including the assessment of their reliability, security, and legal compliance.

Models for creating video from a text prompt; Sora comes from OpenAI, Veo from Google. They are associated with licensing agreements as well as concerns about deepfakes.

An artificially created, realistic digital character capable of interaction that appears to be a real person.

An artificially created digital character that appears on social media as an influencer and has no real physical counterpart.

A technology for the artificial imitation of a specific person's voice; related to the protection of personality rights, fraud risks, and deepfakes.

Rights protecting a person's voice against unauthorized imitation or commercial use.

Protection against the unauthorized imitation of a specific person's voice; related to personality rights as well as the risks of fraud and deepfakes.

A person whose age, health condition, or social situation increases the risk of being negatively affected by an AI system.