… We review different ML solutions to gain safety through each strategy. Note that these solutions
… hand and research debt in AI safety on the other hand, we review and organize practical …

… are of a less consequential nature; examples include streaming services deciding on the
compression level of video packets to transmit to subscribers every few seconds [11], [12], web …

… In Section 3, we present some background on the concepts of safety as well as machine
learning, and ultimately, on safety and machine learning together. Section 4 extensively details …

… aspects of AI (see Section 2), we believe we are far from reaching (1) Inherently Safe AI.
Therefore, we focus on practical machine learning solutions for the two following safety …

… The de facto practical solution is to learn h by empirical risk minimization on the training set
… safety, on the other hand, we review and organize ML solutions into a taxonomy for ML safety…

… of safety has not been formalized in a machine learning context. In this article, we do so by
defining machine learning safety in … In the solution created for this problem, the inherently safe …

… are successfully applied on common benchmark problems as well as on real-world
application problems with very high requirements on the functional safety of the learned solution. …

… in depth and develop solutions. In the present study, we identify, classify, and explore the
open problems in quality assurance of safety-critical machine-learning systems, and their …

… for the machine learning area of reinforcement learning with a focus on NN solutions can be
… methods for reinforcement learning (RL) as developed in [15]; via safe state enforcement for …

… • Using an example deep learning system in a speed control application, how could we
design a simple safety monitoring solution that is feasible for this scenario. Would the monitor …