bit.ly
The Past, Present, and Future of the Semantic Web with Aidan Hogan
In the 45th episode of The AI Digest Podcast, we go back in time to July 4th of 2024 where we spoke with the highly regarded Professor Aidan Hogan and his extensive work focused on the semantic web. Join us as our host, Joaquin Melara, has a friendly chat with Aidan Hogan, Professor at the University of Chile. Aidan also lives a double life as the Deputy Director of the Millennium Institute Foundational Research on Data (IMFD) where they conduct research initiatives with multiple Chilean universities to cross-polinate interdisciplinary ideas. Aidan is also an avid fan and contributor to the Dagstuhl publishing group, focused on championing open-source reports (https://www.dagstuhl.de/en/publishing/series). This episode covers a taste of Aidan’s interest and work in the semantic web and how he has evolved his perspective and contributions. Aidan has branched out to contribute to many different areas in computer science outside of the semantic web since arriving in Chile, including information extraction, graph databases, and machine learning. EPISODE TAKEAWAYS: 1️⃣ - Data as a concept sits inert in 0’s and 1’s. The real question is how to query against that data to get meaningful questions and answers. Database structures facilitate efficient queries. Multitudes of databases and unstructured data found in text or in image then need to be able to interoperate with each other using ontologies to create the ability for richer queries across heterogeneous data sources. 2️⃣ - Capturing knowledge is great, in any format, from paper or stone to digital documents on the web. However to get more value out of knowledge, having a hierarchy of knowledge allows us to extract more localized, contextual information. If we want to integrate knowledge between disparate sources today, we rely on people. By incorporating knowledge models like ontologies, we can encode human knowledge into machines. 3️⃣ - The machine learning area has exploded, there are enough people looking at that area. It is very difficult to be competitive in that, if you are not involved with institutions that have strong computational infrastructure. The theoretical, abstract, and mathematical dimensions of AI still need plenty of research and exploration, in interesting intersections like knowledge graphs and graph neural networks. It does not need to be only statistical or symbolic AI systems, both should have a functional role in future systems. "In terms of the future, it is hard to say… Technologies like LLMs have exceeded my expectations. If you would have proposed something like this 10 years ago, I would have said, ‘no way!’ I would say that LLMs will reach their limits in specific applications due to their unreliable nature. I think we will start to see more composite solutions applying both symbolic and statistical AI technologies in the next few years. " - Aidan Hogan 🎧 Listen on Spotify: https://podcasters.spotify.com/pod/show/the-ai-digest 📺 Watch on YouTube: https://www.youtube.com/playlist?list=PL0aHAh22kpPzhTdZQl-e5LtkN6DsFd707 Follow The AI Digest Podcast for more podcast episodes. If you found this episode informative, be sure to repost. #TheAIDigest #ArtificialIntelligence #bananas ============================================= This podcast episode has been sponsored by the SWARM Community
