Pretrained models are an important term in the fields of artificial intelligence, big data and smart data, and digital transformation. They refer to ready-made, already pre-trained computer models that are used in the development of AI applications.
Instead of starting from scratch every time with huge amounts of data, companies use pre-trained models that have already been „pre-trained“ by experts using large datasets. This saves time, money, and computing power. These models can be used, for example, to automatically analyse texts, recognise images, or even translate speech in real-time.
A vivid example is a medium-sized company that wants to implement a customer service chatbot that can automatically understand and answer customer queries. Instead of „training“ the bot entirely from scratch, the company uses a pre-trained model that already possesses general language understanding. With a small amount of fine-tuning, adapted to its own products and customer questions, the chatbot works quickly and reliably.
In short: Pretrained models make artificial intelligence significantly easier and faster to use – especially for companies that don't have vast amounts of data or large IT teams.
Deutsch 
English (UK)