Waymo unmanned vehicle engineers teamed up with AI researchers from DeepMind (both owned by Alphabet) to customize more accurate and efficient autonomous machine algorithms.
Waymo will be able to easily rebuild control algorithms thanks to the population-based training (PBT) technique that DeepMind has developed to improve video game algorithms. PBT accelerates the artificial evolution of machine code – the algorithm selects the code from the most effective commands. The technique reduces the power of the computer to retrain the neural network and speeds up development by half.
Now the drones operate on the usual transport control code and numerous machine algorithms that determine in real time the road lines, signs, other cars and pedestrians. Each new version of the system is strictly checked before launch.
Google experts believe that the automation of learning AI and the effectiveness of the development …