Interview with Frank van Harmelen on the Future of Generative AI and the Role of Conceptual Modeling

Amsterdam, he has been researching what is now called symbolic AI or knowledge representation.This involves representing the knowledge humans have and use to solve tasks and solve problems.This knowledge is encoded in a form that is also useful for machines, enabling them to use this expert knowledge to solve problems.BISE: Thank you very much for joining us for this interview.What is your perspective on the connection between the BISE community, particularly the business informatics community, and conceptual modeling, and your field of study?Van Harmelen: I think the field of conceptual modeling is a nearest neighbor of knowledge representation.The notion of ontologies for example is a way to structure different types of entities and relationships.There is therefore an overlapping interest between the conceptual modeling field and knowledge representation.I'm a little less familiar with the business informatics field, although some of my colleagues are active there.BISE: You already mentioned some of the historical developments.So how do you perceive the current progress in generative AI particularly from the historical perspective?Van Harmelen: I think we should be honest about this, and the explosion of the capabilities of generative AI has really taken all of us by surprise.Nowadays, if journalists ask me to predict what will happen in five or ten years, I tell them I stopped predicting the future in November 2022, when ChatGPT hit the scene.We had talks in Amsterdam by people from OpenAI and they said they were surprised.From a historical perspective, the transformer model and the underlying algorithms were known since 2017.They were in the open literature; people were playing with them.Apart from some small optimizations, the main step was to put in more compute power and more data.And everybody