AI Augmentation for Business Growth | Managing Partner, RAPYD.AI Consulting | O'Reilly Author | LinkedIn Learning Instructor | Int. Speaker | AI Coach & Trainer
RAG is the future of many business AI applications, but without quality data, it's just a house of cards waiting to collapse. Retrieval-Augmented-Generation (RAG) allows Large Language Models like GPT-4 to quickly access and utilize external knowledge to generate more informed outputs. However, it has its limitations. One critical aspect is data quality. Incomplete, inaccurate, or missing information leads to misleading or irrelevant results, undermining RAG’s effectiveness. We need domain experts to ensure that the data used is accurate, complete, and relevant. Again, that's augmentation. I recently co-authored an article called "RAG (Retrieval Augmented Generation) Architecture for Data Quality Assessment" with Prashanth H Southekal, PhD, MBA and Arun Marar, Ph.D., published on DATAVERSITY. Check out the article here for free: https://lnkd.in/eSjtmJ77 I highly recommend reading it if you're interested in learning more about RAG and its link to data quality.
100% agree! It’s interesting to see some of the output. Cleaned data can still provide inaccurate results. Context and human understanding can get lost in translation to vector storage and similarly search.
Thanks Tobias Zwingmann and Arun Marar, Ph.D. Itws great collaboratring with you on this blog.
Data integrity crucial. Experts needed safeguarding RAG's knowledge base.
Interesting! Thanks for sharing Tobias.
I’m curious with the use case here. Where there are clearly defined “right” answers (“what’s my holiday allowance?”, “what’s my remaining holiday?”), would we not be better off transforming the question in to a direct API call to guarantee correctness and only move to probability if we aren’t aware of a concrete data source? We literally have the right data and are then introducing probability, this seems retrograde. I full support the intent of the article and RAG approach, cleaner, accurate, localised data is the only way enterprises can adopt at scale, I just get concerned that we see AI/GenAI as the solution when there are often simpler ways from A to B!
Completely agree, While we can get good results with smaller models in RAG, the results can be very disappointing even with the most powerful models of the underlying retrieval layer not tuned properly and the major contributing factor to the problem is data-- incomplete data or data that was not handled completely .
Quality means means way more than web scraped. Organizing the web is not enough for accuracy.
The HR database is a great example of how important it is to have good data quality on internal documents that may be used to make some important employee decisions.
That's some valuable insight on RAG and data quality. I'll check out the article for sure. Tobias Zwingmann
Value-Focused, Product-Driven & Outcome-Oriented multi-cultural Practice Delivery Lead and Solutions Architect, known for business and human-centric digital transformations | PRINCE2 Agile® | ITIL® 4
1moA question about RAG, Tobias Zwingmann. I've noticed that when I put information (for example a small document) directly in the prompt for it to action on, I get better results than when I rely on RAG to retrieve the information and action on the whole doc. Now I understand the benefits of RAG and that breaking down the whole into smaller tasks for a better outcome. Wondering if you've noticed this and that RAG vs. direct input needs to be scenario driven.