Learn with NorwAI's experts:
Innovation with generative language models - new course with NorwAI's experts

NorwAI's core team for language models will offer a brand-new course the coming semester.  "Innovation in generative language models" will provide a unique opportunity to learn with the best experts the research team has at its disposal. The course provides insight into how large language models work and enable innovative and intelligent solutions.

The target group for the course will be leaders and decision-makers in innovation processes. Deadline for applications will be February 1st, and the first lecture will take place on February 21st and the day after.

-  The course is free. We will offer the course through NTNU's continuing education program NTNU Videre and registration will take place through NTNU's channels, says Professor Jon Atle Gulla, Director of NorwAI and subject manager.

Pervasive technology

Lectures by Jon Atle Gulla will be given in Norwegian while other lecturers will speak in English. Jon Atle says that dissemination and information are important for Norwegian businesses to be able to use AI. Generative artificial intelligence is a completely pervasive technology that will transform industries and create new products and value chains.

• It reduces costs when it is used to write articles, create summaries and gain an overview of internal databases.
• It automates when translations are done automatically and chatbots replace large parts of customer handling.
• And it generates new products when everyone can generate their own visual expressions and cars can make decisions on their own.
• At the same time, it requires a skills upgrade for the individual worker who currently has few opportunities for further education in this important field.

Digital and physical course sessions

The course “Innovation with Large Language Models” draws on the latest research results from the Norwegian Research Center for AI Innovation (NorwAI) on large language models. The offer is run in a mix of digital and physical sessions:

-  The physical session is used to review key topics in the curriculum, and participants also get to work together on the group assignment. Some lectures and all individual exercises are taught digitally. These are available afterwards, so that they can be reviewed when it suits the participant best. Course participants organize themselves how they want to work on the group assignment, but the result of the group work must be presented at the final lecture in the course, says Jon Atle Gulla. 

Covers multiple sectors

The offer thus reflects the interests of the entire NorwAI and covers sectors such as media, finance, oil and energy, telecom and software development. Exercises will reflect what the consortium sees as important skills in generative AI and language models. 

-  Apart from a physical two-day session, there is no requirement for participation during regular working hours. Exercises are done digitally when it suits them best, and participants organize their own work on group assignments. A good deal of evening and weekend work must be expected to keep up with lectures and exercises, says Jon Atle Gulla. 

Target groups

-   Could you be more specific on the target groups? 

-  Yes, we welcome decision-makers and managers who must understand the structure, potential and limitations of language models. They will need insight into language models to be able to use language models in a responsible and constructive way. There may be participants who will use language models in their own organization, for example journalists and editors, or who are responsible for the development of commercial products.

No prior knowledge of computer technology or artificial intelligence is required, but one should have some understanding of artificial intelligence in general and machine learning in particular. As it is a non-technical subject, programming skills are not required. 

Academic content

• Deep learning
• Generative AI
• Large language models
• Evaluation and benchmarking
• Query systems and search architectures
• Summary generation
• Dialogue systems
• Ethical considerations and responsible AI
• Regulations of AI
• Innovation with large language models

Group exercise

Each group of approximately 5 members is given a topic to explore on their own. A written report of a maximum of 10 pages must be submitted and each group is given 15 minutes to present the results of the group work at the last lecture. Each group is given one of the following topics:

• Non-generative language models.
• Embeddings in language models
• LLMs and knowledge graphs
• Multimodal models
• Sustainability
• "Contextual Awareness" and AGI

More information about the topics and recommended reading material comes in a separate description of the group assignment itself. Students receive a written assessment of the group assignment.

Admission requirements

Qualified applicants will be ranked on a first-come, first-served basis (date of application).

The course will be taught in the spring of 2025:

Five digital sessions (2-3 hours)
Two-day session 20 - 21 February, Trondheim

Curriculum literature/course material

All parts of the curriculum are described in the course material and covered in the lectures.

Individual exercises address key parts of the syllabus. The exercise answers are submitted to the teaching assistant, who approves and reviews the solution with the participants.

The group assignment addresses topics that complement the syllabus, but are not taught in the lectures. This is an assignment that requires somewhat more reflection and discussion among the participants and results in a written report and a presentation.

Other information

The course does not have a final exam. A course certificate is received if the following requirements are met:

• A minimum of two of the three individual assignments are approved
• The group assignment is approved
• Participation in a physical meeting over two days
• The course certificate is issued by the Norwegian Research Center for AI Innovation (NorwAI), NTNU.

Partners

This course is an industry course developed in collaboration with DNB and Schibsted.
Faculty: Faculty of Information Technology and Electrical Engineering

Published 2024-12-17

By: Rolf D. Svendsen