World Model

It makes sense that a clear and logically consistent model of the world allows you to focus on the things that are important to you. It almost acts like a sieve, filtering out the pertinent information.

What happens when we apply this thinking to the GenAI strategies within our organisations? We already know that RAG gives the generative model specific context, leading to more accurate and relevant outputs, perhaps ontology can take this to the next level by providing AI with a type of ‘organisational world model’ as well.

The evolution of AI in strategic games, from Deep Blue to AlphaGo, illustrates the shift from reliance on exhaustive search strategies to adopting neural networks that model the game. This transition underscores the importance of a 'world model' in enhancing the efficiency of search algorithms. As Demis Hassabis states, 'The better your world model is, the more efficient your search can be.'

For organisations, the implications are clear: a robust semantic world model, tailored to the organisation's context and combined with effective search mechanisms over its information, would be game-changing.

Distributed Knowledge Graphs are emerging as a potential solution, facilitating access to data via a structured, organisation-wide ontology. Shared ontologies enable individual applications to align their data with predefined classes, fostering coherence and accessibility across the organisation’s knowledge ecosystem. This is a Semantic Layer in the fullest sense of the word.

Frontier models can integrate with this Semantic Layer, allowing them to adapt their general knowledge to meet your organisation's specific needs. In this sense, ontology provides a framework through which people can teach AIs to understand the unique world model of their organisation.

⭕ Semantic Layer: https://www.knowledge-graph-guys.com/blog/the-semantic-layer
⭕ Core Ontology: https://www.knowledge-graph-guys.com/blog/your-ontological-core
⭕ Yann LeCun World Models Paper: https://arxiv.org/pdf/2403.00504
⭕ Embrace Complexity Conclusion: https://medium.com/@Tonyseale/embrace-complexity-conclusion-fb8be6f39deb