I have just returned from EFSA's Data Readiness for AI Symposium and going deep on semantics, ontologies and knowledge graphs.
As artificial intelligence advances, the prevalent use of knowledge graphs (KGs) and retrieval-augmented generation (RAG) has become a much hyped cornerstone for structuring information and ensuring accuracy. However, these methods may also be holding AI back.
By aligning AI systems with traditional human ontologies — which are structured semantics frameworks grounded in human experience and logic — we may inadvertently limit AI's potential to discover new, purely correlation-based patterns that defy our rigid classifications.
The Constraints of Human Ontologies
Human ontologies are, by nature, restricted to what we already understand. Through scientific method, and human consensus seeking (which is less scientific then we might be comfortable with), we arrive at ontological truths.
KGs and RAG reinforce these structures by embedding human logic and categories into AI's operational framework. This grounding has clear benefits for accuracy and interpretability; e.g. showing us new knowledge from within the box of what we accept as truth. However, it risks filtering out patterns or relationships that do not align with established categories. AI's inherent power lies in its ability to process vast amounts of data, beyond our biases or perspectives. Forcing it to operate within predefined human categories constrains this ability, keeping it from recognizing connections that are opaque or invisible to us.
Let me illustrate with a simple example. Existing knowledge graphs in food safety demonstrate the ontological path from flavour, through compounds (molecules) to human health impact.
That makes sense. It's scientific. But by grounding Machine Learning (ML) to the ontology, we shutout pattern seeking between flavour and health. Now bear with me, I am not implying that there may be some new age non-scientific connection between flavour and health. That would not be scientific. But it is not absurd to speculate if flavour and the human experience of taste have complex interactions within the psychological and psychosomatic reality. Even though embodiment is a well-studied and scientifically accepted discourse, by limiting the AI to traditional human ontologies through KGs and Graphs we are severely scoping pattern seeking within the available data sets.
Embracing Machine Ontologies Through Correlation-Based Discovery
Imagine AI models set free from rigid ontological frameworks and instead driven purely by mathematical correlation. Rather than forcing AI to follow existing human ontologies, we could allow it to generate machine ontologies — frameworks of knowledge that reflect patterns in the data itself. Such machine ontologies would not make sense immediately but could reveal connections we hadn't anticipated. Only ex-post facto, by mapping AI's conclusions back into our human frameworks, would these new ontologies gain meaning.
Why Machine Ontologies Could Be the Future of Knowledge Discovery
Allowing AI to build its own ontology based on correlation-driven patterns opens up a new frontier in knowledge discovery. AI could uncover structures within complex data that reveal hidden relationships, enabling breakthroughs that go beyond what structured human-driven frameworks allow. This approach could lead to innovations in fields such as scientific research, where many causal relationships are unknown, and in complex systems where human-defined categories obscure deeper truths.
The path forward, then, is not necessarily to discard human ontologies but to treat them as one perspective among many. By enabling AI to build knowledge frameworks beyond our own, we might unlock a realm of insights and novel patterns that fundamentally reshape our understanding, leaving KGs and RAG as supportive tools rather than dominant paradigms.
Hallucination and Explainability — Rethinking the Flaws
Hallucination, within the space of Generative AI, is increasingly seen as a flaw in machine intelligence. Lack of explainability, a profoundly human characteristic, is seen as one of the major (and now regulated) risks for the use of AI. Rather than flaws, these are indicators of AI's potential to transcend our rigid structures and explore uncharted territories of knowledge. Yes, most of the time, in these very early days of the AI and big data revolution, we can show that they are absolutely, scientifically, just wrong. But these early hallucinations in generative AI might, in the near future, signal unexplored connections or latent patterns within vast datasets, presenting opportunities for novel discovery rather than error alone.
Our preoccupation with explainability reflects a human desire for control and comprehension, but if we limit AI to only what we can explain, we restrict its ability to reveal insights beyond our own conceptual boundaries. Embracing machine ontologies, where patterns emerge from correlations rather than pre-existing semantic frameworks, could unlock knowledge that becomes meaningful to us only retrospectively. This approach doesn't eliminate the need for responsible AI; it expands our approach, acknowledging that AI's power lies in being freed from human semantics to unveil new dimensions of knowledge and progress.
The Takeaway
In short, if we allow AI to operate freely, building knowledge based on correlations and emergent ontologies, we invite a collaborative evolution of knowledge, where human and machine insights coalesce to redefine what we consider "true" or "known." This will challenge our frameworks but may ultimately advance our grasp of complex systems in ways we can barely imagine. Ontology is not dead — but its role must evolve, from gatekeeper of truth to co-creator of knowledge.