How do I build my own language model? Fine-tuning a Large Language Model (LLM)

Large Language Models (LLM), such as those behind ChatGPT, are transforming technical communication. There are many use cases for AI, from text generation to voice assistants. However, despite their power, generic LLMs sometimes do not provide the right solution for a company's specific projects. Generic AIs have a lot of general knowledge and can handle many languages and general texts – but they often struggle with requests from specific domains.

Is it possible to train a generative AI in a corporate language and to train it with the knowledge of the company?

We have trained and fine-tuned an LLM with technical documentation in two languages and examined the effects. In our presentation, we show the selection criteria for an LLM, the process of fine-tuning the LLM, and the results of the company-specific LLM.

What you will learn:

• What are Large Language Models?
• Use of language models in technical communication
• Fine-tuning a large language model: process and results
• Outlook for productive use in technical communication

The video is a recording of the presentation given by Helle Hannken-Illjes and Ulrike Parson at the tekom annual conference 2023. The presentation is in German.

We advise on the use of artificial intelligence, for example through the use of language models such as ChatGPT, in technical communication and for the various target-group- oriented channels. Find out more!