Personal AI chatbot: A workshop for visionaries

In October, the Kufstein University became the setting for an exciting deep dive into the world of artificial intelligence: under the direction of Dr. Thomas Ströhle, an expert in data science, participants immersed themselves in the development of personalized chatbots. The focus was on the innovative technologies Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG).

AN INTERACTIVE CONCEPT WITH A FOCUS ON PRACTICE

What is it all about? Ströhle explains: “LLMs are AI models that have been trained on large text data sets to understand human language and create texts. They are used in applications such as text generation, translation and customer service.” RAG systems go one step further and combine LLMs with search techniques to retrieve relevant information from external sources as needed. This enables them to provide more precise answers that go beyond the model knowledge. “So you combine internal company knowledge with AI,” explains Ströhle.

The workshop began by introducing LLMs and examining their capabilities and applications. The overarching question was: how do you communicate efficiently with LLMs? Participants practiced optimizing their inputs to achieve the best results.

After that, the participants learned more about the application of RAG systems. Here, their own systems were built and enriched with data to generate context-based answers.

AI FOR EVERYDAY LIFE AND PROFESSIONAL LIFE

And where could these personalized chatbots be used in practice? "For example, a personalized chatbot with RAG System could automatically process frequent technical queries in IT support, such as resetting passwords, diagnosing errors or providing instructions for installing software," explains Ströhle. By integrating it into the company system, it could provide information tailored to individual users - which would speed up and relieve the burden on support.

But there are also possible applications in the field of e-learning: “A personalized chatbot could provide individual learning tips, quiz questions or explanations tailored to the user's level of knowledge and track learning progress,” explains Ströhle. As a result, the learning process is made more efficient and motivating.

Ströhle found inspiration for the workshop in his current research project, in which an IT chatbot is being developed for over 50,000 employees.

Ströhle emphasizes: “LLMs and RAG systems are increasingly playing a central role in data science education. Students not only learn modeling, but also how to integrate dynamic knowledge sources to create intelligent systems, as they offer new approaches to analyzing and processing text data.” Students are now learning not only to train models, but also to combine their use with dynamic knowledge sources to obtain more precise and context-based answers. This shifts the focus from pure modeling to the implementation of intelligent systems that can flexibly access external information.

INSPIRING INSIGHTS

In addition to practical exercises and lively discussions, the workshop participants were surprised by an eye-opening moment: the evaluation of RAG systems can also be carried out with another LLM. In plain language, this means that the AI can thus initially evaluate itself before we humans assess it.

SUBSIDIZED BY AI4GREEN

Supported by the Interreg project AI4GREEN, the workshop impressively demonstrated how technologies like these can contribute to the sustainable and future-oriented use of AI.

An inspiring day that offered participants new perspectives on personalized AI solutions and strengthened them for the coming challenges of digitalization!

Links:

• Bachelor's degree program in Drone Engineering & AI-based Innovation (full-time)
• Master's degree program in Data Science & Intelligent Analytics (full-time)
• Master's degree program in Data Science & Intelligent Analytics (part-time)