LLM and Local Knowledge
| MediaWiki Users and Developers Conference Fall 2024 | |
|---|---|
| LLM and Local Knowledge | |
| Talk details | |
| Description: | How does Generative AI integrate with the framework of Digital Humanism? This talk highlights the potential of AI technologies while addressing ethical concerns like bias and misinformation. |
| Speaker(s): | Julia Neidhardt, Thomas Kolb |
| Slides: | see here |
| Type: | Talk |
| Audience: | Everyone |
| Event start: | 2024/11/06 10:30:00 |
| Event finish: | 2024/11/06 11:00:00 |
| Length: | 30 minutes |
| Video: | click here |
| Keywords: | |
| Give feedback | |
This presentation explores the integration of Generative AI within the framework of Digital Humanism. It highlights the potential of AI technologies, such as large language models, for creating human-centered innovations while addressing ethical concerns like bias and misinformation. The focus is on practical applications, including enhancing local knowledge platforms like the Vienna History Wiki, and discussing emerging topics such as multi-agent systems. The presentation emphasizes balancing technological advancements with human values and societal needs to foster a better, ethically aligned future.
Contents
Detailed Description[edit]
This presentation focuses on the role of Generative AI in the context of Digital Humanism, with a practical example of its application to the Vienna History Wiki. Presented by Assistant Professor Julia Neidhardt and PreDoc researcher and PhD student Thomas Kolb from TU Wien, it provides an overview of how AI technologies can be used effectively while ensuring they align with ethical standards and societal needs.
Key Areas Covered[edit]
Introduction to Generative AI[edit]
- Generative AI uses machine learning models to create content such as text, images, and music by identifying and imitating patterns in large datasets.
- The presentation highlights practical applications and challenges, such as biases, misinformation, and copyright issues.
Digital Humanism Framework[edit]
- Digital Humanism emphasizes the development and use of technology that respects human rights and societal values.
- The talk explores how AI can support societal needs while ensuring ethical integrity.
Practical Use Case – Vienna History Wiki[edit]
- Demonstrates how AI tools can enhance the Vienna History Wiki, a platform for local knowledge.
- Specific methods include:
- Retrieval-Augmented Generation (RAG): Combining local data with large language models.
- Data retrieval techniques such as semantic data access via SPARQL, MediaWiki Action API, and embedding-based approaches.
- Prompting methods highlighted include chain-of-thought reasoning, in-context learning, and parameter-efficient fine-tuning (PEFT).
Emerging Topics and Technologies[edit]
- Discussion of multi-agent systems that combine various retrieval and analysis strategies to improve efficiency and outcomes.
- Considerations on computational requirements, scalability, and model optimization.
Challenges and Key Insights[edit]
- Risks addressed include hallucinations, incomplete retrievals, and biases in AI-generated outputs.
- Strategies for reducing risks include implementing robust retrieval systems and promoting transparency in AI operations.
Future Outlook[edit]
- Insights into potential applications and improvement areas for AI systems.
- Emphasis on using ethical frameworks like Digital Humanism to guide AI development toward human-centered innovation.
Summary[edit]
The presentation provides insights for researchers, developers, and industry working with AI. It offers guidance on using AI’s potential while addressing ethical challenges, ensuring that technological advancements remain focused on societal and human values.
