07.05.17 - 12.05.17, Seminar 17192

Human-Like Neural-Symbolic Computing

Diese Seminarbeschreibung wurde vor dem Seminar auf unseren Webseiten veröffentlicht und bei der Einladung zum Seminar verwendet.


Human-like computing is a new area of research that seeks to incorporate into computer science aspects of how humans learn, reason and compute, whilst recognising the importance of the recent trends in big data. Data science methods and techniques have achieved industrial relevance in a number of areas, from retail to health, by obtaining insight from large data collections. Notably, neural networks have been successful and efficient at large-scale language modelling, speech recognition, image, video and sensor data analysis. Human-beings on the other hand are excellent at learning from very few data examples, capable of articulating explanations and resolving inconsistencies through communication.

We argue that data science requires an ability to explain its insights. As a case in point, despite the success of neural networks at performing language modelling, very little progress has been made at understanding the principles and mechanisms underlying language processing. Techniques of knowledge extraction ought to be investigated and applied in the context of data science success stories to enable a systematic study and the data-driven formulation of sound theories, capable of explaining the insights and moving the research forward. In addition to knowledge extraction models and algorithms, various aspects of human-machine and human-agent interaction and communication will be fundamental to the process of achieving human-level understanding and better scientific theories, as well as systems capable of reasoning about what has been learned.

In this Dagstuhl Seminar, we will bring together world-leading researchers and rising stars in the areas of neural computation, language modelling, artificial intelligence (AI), knowledge extraction, computational logic, machine learning, neural-symbolic computing, cognitive psychology, cognitive science and human-computer interaction to discuss, investigate and formulate the requirements and fundamental applications of human-like computing.

Language modelling tasks will be considered as a first case study, and the methodology of neural-symbolic computation will serve to underpin the discussions on knowledge extraction, representation and learning, and as baseline for comparison with alternative approaches, such as statistical relational AI.

Specifically, this Dagstuhl Seminar seeks to produce (i) better bridges between symbolic and sub-symbolic reasoning and learning, and between big data and human-like learning; (ii) comparative analyses and evaluations of the explanatory capacity of language modelling tools and techniques; (iii) designs and applications of knowledge extraction methods and techniques towards life-long learning and transfer learning between areas of application.

Expected outcomes include a roadmap towards human-like computing, a manifesto in a journal special volume, and the definition of a human-like neural-symbolic challenge and evaluation framework.

Creative Commons BY 3.0 Unported license
Tarek R. Besold, Artur d'Avila Garcez, Ramanathan V. Guha, and Luis C. Lamb