The Art of the Possible with Graphs
0 likes214 views
Dr. Jim Webber discusses the transformation in data management due to the emergence of connected data and the advantages of graph databases over traditional relational databases. He highlights how graph databases enable deeper insights by uncovering relationships in data, driving innovation and improving operational efficiency through advanced analytics and machine learning. The presentation emphasizes the need for modern solutions like knowledge graphs and AI integration to stay ahead in a rapidly evolving data landscape.
![LLMs hallucinate in
interesting and
potentially troubling
ways
[Yejin Choi et al, U Washington]](https://image.slidesharecdn.com/jw-theartofthepossiblewithgraph-231030142216-263c60b9/85/The-Art-of-the-Possible-with-Graphs-44-320.jpg)
![LLMs hallucinate by in
interesting and
potentially troubling
ways
[Yejin Choi et al, U Washington]
Knowledge model 435x smaller and 16% better than raw LLM](https://image.slidesharecdn.com/jw-theartofthepossiblewithgraph-231030142216-263c60b9/85/The-Art-of-the-Possible-with-Graphs-45-320.jpg)
![New techniques
promise a revolution
in creating better and
smaller LLMs
[Yejin Choi et al, U Washington]](https://image.slidesharecdn.com/jw-theartofthepossiblewithgraph-231030142216-263c60b9/85/The-Art-of-the-Possible-with-Graphs-46-320.jpg)
The Art of the Possible with Graphs
- 1. The Art of the Possible Dr. Jim Webber Chief Scientist, Neo4j
- 13. �� �� ��
- 14. Legacy approaches are not optimized to handle the shift to connected data
- 15. Imagine you have to “normalize” this data into a relational database with SQL
- 16. Previous data technology can’t tackle today’s challenges
- 17. The world is changing
- 18. The world is changing As things get more connected, data does too
- 19. The world is changing Knowledge, not just data, is a competitive advantage
- 20. Gen AI Fraud Supply Chains Medicine Space Travel
- 21. Gen AI Fraud Supply Chains Medicine Space Travel The insights are latent in your data’s connections
- 22. Index-Free Adjacency: Fast, Flexible, Scalable Relational Database Person Person-Friend Friend Native Graph Database
- 23. Graph databases natively store and query relationships Data model purpose-built for connected data Balance of schema flexibility and structure Powerful and productive query language
- 24. Connectedness and Size of Data Set Response Time Relational and other NoSQL databases Native Graph Database 1000x Advantage 5+ hops 3+ degrees Thousands of connections 0 to 2 hops 0 to 3 degrees Few connections Performance Advantage of Graph as # of Joins Increase From Minutes to Milliseconds
- 25. Uncovering relationships in connected data is key to deeper understanding of Context Behavior Preferences
- 26. Uncovering relationships in connected data is driving game-changing innovation
- 27. Enabling extraordinary outcomes Powering applications that are impossible with other technologies
- 28. Graph Transactions, Storage & Querying Graph Analytics, ML, & Data Science Intelligent Operational Systems Better Predictions for Analytics Neo4j Differentiation Neo4j Database built for Operational and Analytical Workloads
- 29. Graph Applications (System of Record Applications) Application Stack Real-time Messaging & Processing (Clickstream, IOT, CDC etc) Developer Intelligent apps Data Analyst Query and analyze Biz Analyst Visual Analytics Data Scientist Algorithms and features Data engineer Get clean, useful data ML Engineer ML Ops Data Platform (Data Platforms-Snowflake, BigQuery, Databricks) AI/ML Ops BI Platforms Graph Analytics & Analytical Apps (Real-time applications, visualizations and algorithms) Powered by Market Trends Modern Data Platform
- 30. ChatGPT is just ONE Year Old!
- 32. + Combine explicit and implicit relationships for reliable and contextual answers LLMs for language generation Retrieval Augmented Generation Knowledge Graph Construction
- 33. Neo4j Inc. All rights reserved 2023 LLMs for language generation
- 34. Generate personalized Natural Language experiences Prompt (Description of desired text with personalized data) Response (Generated personalized Text) LLM API User Experience (e.g. Email, custom Ad) Enterprise App (gets context from KG) Knowledge graph LLMs for language generation
- 36. User Embedding models LLMs Operational database Prompt (Description of desired text with personalized data) GenAI app Grounding with Retrieval Augmented Generation (RAG) Retrieval Augmented Generation
- 38. Natural Language Search on unstructured data using implicit relationship Prompt + Relevant Information LLM Embeddings API LLM Chat API User Vector Search Prompt Response Relevant Results / Documents Embedding 2 3 1 Vector Index Retrieval Augmented Generation
- 39. Grounding LLM Responses with Implicit and Explicit Search Through Knowledge Graph User Question Vector Vector Index LLM (Embedding Model) Application LLM Smart Search Explicit and Implicit Results Similarity Search Implicit Results Explicit results in a Neo4j knowledge graph Natural Language Results Retrieval Augmented Generation
- 41. LLM Accuracy Vector-Only RAG Use Case Maturity (time) Vector + Graph RAG Business value unlocked with graphs Use Case Effectiveness
- 42. This is just the beginning The fusion of AI and graphs will be amazing
- 43. This is just the beginning Researchers are leading the way
- 44. LLMs hallucinate in interesting and potentially troubling ways [Yejin Choi et al, U Washington]
- 45. LLMs hallucinate by in interesting and potentially troubling ways [Yejin Choi et al, U Washington] Knowledge model 435x smaller and 16% better than raw LLM
- 46. New techniques promise a revolution in creating better and smaller LLMs [Yejin Choi et al, U Washington]
- 47. The future is connected
- 48. The future is connected Rethink what’s possible
- 49. 250K data pros Join our Global Community Enroll in GraphAcademy 20K completed courses Attend NODES 2023 24 hours, 10K+ attendees OCT 26th
- 50. Thank you Dr. Jim Webber jim.webber@neo4j.com