|Beyond Moore’s Law and Rebooting Computing with Probabilistic Machines, a talk by Prof. Jorge Lobo|
|Ciência 2016, Lisbon
We propose Probabilistic Machines that can make a difference for robots that operate in an uncertain world. Moore’s Law allowed us to have smart robots, but uncertainty and incompleteness in models and systems used in robotics lead to heavy computations, requiring power hungry heavy machines. If we look at nature and biology, things can be done in other ways. We revisit the correctness contract between hardware and software, and propose Bayesian probabilistic machines using stochastic computing. The BAMBI project is presented BAMBI (EU FET project, Bottom-up Approaches to Machines dedicated to Bayesian Inference), showing recent results that enable the final goal of having efficient robots dwelling in uncertainty but dealing with it!