Looking for community ideas for the 2026 Ontology Summit
Gary Berg-Cross
I thought I would reach out to the community for advice and thoughts on suitable topics & speakers for this Summit. Regular readers of the forum may have noticed for example that John Sower has started some initial discussions on ideas that he would suggest would be part of it.
A good part of the committee is thinking of picking some particular past summits and organizing talks on that theme to provide some historical context but also affording an update of what's new or what's changed.
For my part, based on my prior participation and interests, I'm considering the obvious topic of the relationships between machine learning and ontologies. This is broadly a topic that I've Co chaired in the past.
For example the OntologySummit2017 on "AI, Learning, Reasoning, and Ontologies" included a track on:
Using Automation and ML to Extract Knowledge and Improve Ontologies
We also had:
Estevam Hruschka: Overview of Never Ending Language Learner NELL and
Alessandro Oltramari: From machines that learn to Machines that Know!
The Challenges of Hand-Crafted Knowledge (as Richard Sutton puts it):
We might update these since ontology engineering is still an iterative and spotty (non-uniform progress in its activities and process) & there are obvious bottlenecks and obstructions.
So we re-look at how Machine Learning (ML) can help generate semantic resources like KB to help Develop Ontologies. Now the knowledge available in transformer-based systems provides additional help to reduce noisy data to further the quality of ontologies. See for example Chris Mungall’s summary https://monarchinit.medium.com/exploring-the-power-of-ai-in-ontology-development-and-curation-highlights-from-the-obo-academy-7f0aa318c9d7
So one topic broad topic is what types of machine learning can help develop ontologies and improve their quality. There is, of course, the other direction about how formal knowledge resources like ontologies can be used to improve machine learning related applications this was a topic that people like Gary Marcus talked about in the OntologySummit 2024 on “Techniques for and with Ontologies and Knowledge Graphs" and we could have an update on that topic of hybrid systems that was covered there.
Here are few, perhaps overlapping, topics that might update what we have covered before (and I’d like people’s thoughts on these):
1. Agents - One area developing rapidly now uses the concept of agent-oriented AI (i.e. meaning creating intelligent systems that can act autonomously within their environments, making decisions, learning, and adapting based on interactions). This approach is higher level & contrasts with the now "traditional methods", such as Transformers, which primarily focus on processing and generating data based on patterns learned from large datasets.
We could have a discussion of agent based/goals oriented systems that go beyond transformers.
2. Reinforcement agent-like learning.
In contrast to transformers there is also work on reinforcement agent-like learning -the beauty of RL agents lies in how naturally they learn. Richard Sutton’s work- Essentially, a computer program is not told what actions to take in reinforcement learning, but is given a certain set of rewards to pursue. It must then discover which actions yield the most reward by trying them out, thus enabling experiential learning. (for more see: https://www.ibm.com/think/news/turing-award-winner-on-agi)
3. Update on our 2024 Summit topic on hybrids -cognitive architectures to combine reasoning and learning as similar objectives outlined by Gary Marcus (see also https://www.sciencedirect.com/science/article/pii/S2667305325000675)
4. Finally there is also the area of predictive architectures. Yann LeCun, of deep learning fame, has proposed several predictive AI architectures that aim to enhance the capabilities of machines to understand and anticipate future events based on past data. For example, Joint Embedding Predictive Architecture (JEPA), isa non-generative AI model that learns by predicting parts of a video or image that have been masked. This work emphasizes the importance of predictive coding and self-supervised learning as foundational elements in AI development. We might try to get him to provide a view of how such (more cognitive) architectures may bridge between transformers and better reasoning.
Thoughts and comments?
Marco Neumann
--
All contributions to this forum are covered by an open-source license.
For information about the wiki, the license, and how to subscribe or
unsubscribe to the forum, see http://ontologforum.org/info
---
You received this message because you are subscribed to the Google Groups "ontolog-forum" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ontolog-foru...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CAMhe4f0dx0aFtUPOxJnGyWdS%2BPUz6F_ecieKXrJTwGrLKMpFyg%40mail.gmail.com.
---
Marco Neumann
Gary Berg-Cross
To view this discussion visit https://groups.google.com/d/msgid/ontolog-forum/CABWJn4SPnkrNDh3_V83UbY_7LBfPHrg%2Bp51%2BF22rQvEOJMNjoQ%40mail.gmail.com.