Rapid advances in artificial intelligence (AI) and machine learning are creating products and services with the potential not only to change the environment in which actuaries operate, but also to provide new opportunities within actuarial science. These advances are based on a modern approach to designing, fitting and applying neural networks, generally referred to as “Deep Learning”. This paper investigates how actuarial science may adapt and evolve in the coming years to incorporate these new techniques and methodologies. Part 1 of this paper provides background on machine learning and deep learning, as well as an heuristic for where actuaries might benefit from applying these techniques. Part 2 of the paper then surveys emerging applications of AI in actuarial science, with examples from mortality modelling, claims reserving, non-life pricing and telematics. For some of the examples, code has been provided on GitHub so that the interested reader can experiment with these techniques for themselves. Part 2 concludes with an outlook on the potential for actuaries to integrate deep learning into their activities. Finally, a supplementary appendix discusses further resources providing more in-depth background on machine learning and deep learning.

Rapid advances in artificial intelligence (AI) and machine learning are creating products and services with the potential not only to change the environment in which actuaries operate but also to provide new opportunities within actuarial science. These advances are based on a modern approach to designing, fitting and applying neural networks, generally referred to as “Deep Learning.” This paper investigates how actuarial science may adapt and evolve in the coming years to incorporate these new techniques and methodologies. Part 1 of this paper provides background on machine learning and deep learning, as well as an heuristic for where actuaries might benefit from applying these techniques. Part 2 of the paper then surveys emerging applications of AI in actuarial science, with examples from mortality modelling, claims reserving, non-life pricing and telematics. For some of the examples, code has been provided on GitHub so that the interested reader can experiment with these techniques for themselves. Part 2 concludes with an outlook on the potential for actuaries to integrate deep learning into their activities. Finally, a supplementary appendix discusses further resources providing more in-depth background on machine learning and deep learning.

In this study, an inverse Real Time Kinematic (RTK) GPS positioning approach is presented and discussed. GPS data from a remote receiver mounted on an unmanned vehicle is sent to the base station to determine the vehicle location at the centimetre level of accuracy in real time. Control of the vehicle's movement and manoeuvring can thus take place at the base station, which could be a few kilometres away, and the vehicle can be sent to specific locations to do certain tasks. Two GPS antennas were used in order to increase positioning reliability and estimate real time heading and pitch of the vehicle for better control. In addition, the vehicle was fitted with cameras with wireless video transmission to provide the operator with a good level of vision in different directions during operation. The main applications of such a system would be in exploration, breaching and clearance of minefields, and hazardous situations such as fire fighting of burning oil wells. The system design and techniques employed are discussed first. Issues addressed include antenna layout, data transmission, solution algorithm, and ambiguity resolution. The proposed system was mounted on a prototype vehicle and tested under different satellite visibility conditions. Results show that high positioning accuracy with reliable ambiguity resolution can be obtained with the developed approach if the number of observed satellites is 5 or more and PDOP is less than 5. Heading and pitch were determined within 0·2–0·3 degree using a 1·2 m long baseline. This performance can be improved as the length of the on-board baseline increases.

Navigating in unfamiliar road environments is a common and demanding cognitive activity. If this cannot be accomplished successfully, there are implications for increased driver workload, delays due to navigation errors, potentially unsafe road behaviour such as late lane changes, and inappropriate traffic management. To enable successful navigation through such environments, it is necessary to understand what the navigation task entails, and what a driver's corresponding information requirements are. This paper reports the results of a study that sought to identify what information is used by drivers when navigating within a range of urban driving environments, how this information is used, and what the implications are for the design of navigation aids.