DevGott001
commented
1 year ago What are the best ways to deploy such models
Open SamRepository opened 1 year ago
DevGott001
commented
1 year ago What are the best ways to deploy such models
Hi AMAI Team,
I wanted to start by commending you on the great work you have produced. However, it appears m to me that certain aspects and technologies crucial to important AI subfields like Symbolic/Neuro-Symbolic AI and Knowledge Engineering have been missed. These areas of study are becoming increasingly important in the development of AI systems, as they focus on creating knowledge-based systems that can reason and learn from symbolic representations of information.
Symbolic AI involves the use of logical and linguistic structures, such as description logics and NLP techniques, to represent knowledge and manipulate it to solve problems. Unlike recent AI approaches that rely on statistical methods and machine learning, incorporating symbolic AI techniques enables AI systems to reason logically and make inferences based on the rules and relationships within the knowledge base.
In the case of Neuro-Symbolic AI, symbolic reasoning and neural network-based learning are integrated at multiple levels, including the use of embeddings and graphs techniques to combine neural networks with rule-based systems, incorporating symbolic reasoning into deep learning models, and using neural networks to learn the structure of symbolic knowledge. The aim is to create AI systems that can reason about complex, abstract concepts and learn from experience in an efficient and interpretable manner. Moreover, this direction can provide answers for more explainable AI systems and act as a bridge between big AI subfields.
Knowledge Engineering, on the other hand, is focused on creating and managing knowledge in AI systems. This involves developing methods to represent knowledge in machine-readable forms such as RDF, OWL ontologies, and knowledge graphs (Semantic Web Stack), as well as using or creating tools and techniques to manage, maintain, and update this knowledge over time, such as RDF stores, SPARQL endpoints, and graph databases. By applying knowledge engineering principles, AI systems can better leverage the knowledge they have acquired and make more informed decisions.
While the work you have presented is undoubtedly impressive, incorporating principles and technologies from these subfields into your roadmap would help ensure a more complete picture for AI systems that can capture, learn, adapt and reason. I hope this feedback proves useful for your future endeavors.
Regards, Samir"