While our focus at the moment is firmly on city bikes, driverless cars – and what they mean for shared transport services in future cities – are really not that far away. An important requirement here is dedicated chips for running the ML algorithms used in self driving systems. These need to cope with high velocity data, like that coming from high-definition cameras, in real time. You don’t want to collide with a lamp post due to chip latency issues. Chip makers look to be taking some significant steps forward here, lead by Nvidia and following on from its lead in the GPU space for training ML systems. Reuters has a good overview – definitely worth reading.