Graph-based artificial intelligence uses knowledge graphs to provide context and connections between data, enhancing AI models. Knowledge graphs allow for graph-based feature extraction and graph-assisted learning. They provide a system of record for building out graph AI and processing engine for carrying out graph-enhanced AI. Graphs can enrich AI inputs through global and transactional graph algorithms, revealing relationships and causality for applications like predictive modeling. Overall, incorporating graphs can lead to faster, more accurate AI development by operationalizing real-time analytics and monitoring on the graph system of record.
5. Has many dependencies, including observed behavior & human training
Breaking Down AI
Algorithms that learn & improve over time
New* and Valuable
*Static algorithms whose results improve as the data improves are not new
6. Breaking Down AI
here be graphs!
Connection-centered intelligence
reveals context & causality
New and
Valuable
7. Actually: there’s nothing magical here.
It’s just about tracing the relationships.
Lots of terms for this…
e.g. “Inferencing”: Tying Remote Causes to Proximate Effects
8. Applications of Graphs in
“Computer-Based Decision Making”*:
+ “Graph-Based
Algorithms”
+ “Graph-Assisted
Learning”
+ +
“Machine
Intelligence”*Or if you prefer… AI
9. You might be wondering….
Neural Networks & Decision Trees
Are ___________?
10. As they are inside of the ML black box,
the graphs inside probably* don’t matter much
in the application of AI
16. MATCH path = (:Animal {Entity:"Wolves"})-[*]->(:Landscape {Entity:"Rivers"})
WITH extract(node IN nodes(path) | node.Yellowstone) AS factor, rand() AS number
RETURN factor AS How_Wolves_Affect_RiverStability
ORDER BY number
LIMIT 5
Yellowstone National Park Ecosystem
Step 3: Query! Trophic Cascades Example
Conclusion:
19. Graph-Assisted Learning. Example #1 of 2:
Graph-Based Feature Extraction
Further reading: https://neo4j.com/blog/machine-learning-graphs-fake-news-epidemic-part-2/
General idea:
the input of your ML algorithm is derived from a graph query
MATCH p = (a1:Article {title: 'The Fake News Epidemic'}) -
[MENTIONS]->(n:Topic:Entity)<-[MENTIONS]-(a2:Article)
WITH count(p) AS commonality, a2.article_id WHERE
commonality >=2 RETURN a2
20. Graph-Assisted Learning. Example #2 of 2:
Knowledge Graphs
Further listening: 1a16z Podcast: The Taxonomy of Collective Knowledge
“A lot of the things called AI are just fancy ontologies1”
22. Machine Intelligence Example
eBay Shopbot: Conversational Commerce
Further reading: https://medium.com/@rjpittman/cracking-the-code-on-conversational-commerce-775b5172f312
Medium Post by RJ Pittman: Cracking the Code on Conversational Commerce
25. The Result:
#1: System of Record for Building Out Graph AI
#2: Processing Engine For Carrying Out Graph
26.
27. Summary: Graph Boosted Artificial Intelligence
Knowledge Graphs
Provide Rich
Context for AI
AI Visibility
Human-Friendly
Graph Visualization
Graph Enhanced AI Models
Faster, More
Accurate Development
Graph Execution of AI
Operationalize Real-Time
OLAP and Monitoring
Graph Analytics
Enrich AI Inputs with
Graph Algorithms
Graph System of Record
Maintain a Source of
Connected AI Truth