AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
The Challenge of Building Deep Learning Approaches
The hardness of building deep learning approaches that are very good at relational reasoning in the logical sense is akin to the challenge of building language interpretation. We have better tools coming from the logic world, from the logic in computer science school. So what we do in neuro symbolic AI or neuro symbolic computing is not to throw all the data at the same point and feed a huge neural network with an even larger amount of data. What we do is we do some pre-processing, we provide knowledge representation tools to proper design a learning system where symbolic structure, discrete structures are first learned by this machine learning mechanism. And then you verify then you do your cross validation, then you use your statistical approach to validate