Manish Sharma is the co-founder of Resolve AI. With a rich history spanning over two decades in the technology industry, Manish offers profound perspectives on the intersection of artificial intelligence, information architecture, and knowledge management.
Tune in as Manish dissects popular AI myths, underscores the importance of bridging the technological gap, and emphasizes the need for robust security measures in today's digital landscape.
Key takeaways:
- When implementing AI solutions like large language models, CISOs should ask questions around data security, access controls, model guarantees, and emerging risks like prompt hacking to properly manage risks.
- Information architecture is critical for data privacy, security, and ensuring AI systems can only access appropriate data sources and provide the right information to different user groups.
- Retrieval augmented generation using a knowledge graph or index is important to avoid hallucinations and ensure AI systems can only respond based on curated data sources.
- Scripted responses may be needed in some cases like legal to provide verbatim answers instead of generated responses.
- User personas and metadata are important to ensure AI systems understand the context and privileges of different user groups to provide appropriate and non-confusing information.
- When integrating AI solutions with knowledge repositories like SharePoint, only curated subsets should be connected instead of entire repositories, and information should be properly tagged and structured.
Quote of the show:
"A key to successful AI integration is not just in understanding the technology itself but in grasping the nuances of user needs, processes, content, and knowledge that remains timeless, no matter the advancements in tech. Coming to grips with that is where the real value lies."
- Manish Sharma
Links:
Ways to Tune In:
Thanks to our sponsors: