Automated testing aims to generate a high coverage suite of tests which can identify defects in a unit under test. Works in automated testing have achieved successes in regression testing and finding safety property violations. However, we are far from the dream of an automatically generated comprehensive bug finding test suite. Automated testing tools struggle to uncover defects in benchmarks of real world known bugs. In this talk I will identify a number of shortcomings in automated testing for bug finding and propose the potential for deep learning to bridge this gap.