Many thanks to MEMS Journal for inviting us to present at the 2017 Automotive Sensors and Electronics Conference. We don’t believe the marketing hype that training a deep neural network running on an automotive supercomputer will deliver reliable, safe, mainstream autonomous vehicles. So we presented the counter case: pushing intelligence to the sensors and fusing classifiers, not raw data, on a substantially less powerful central ECU. Rather than cries of heresy from the audience of OEM and tier supplier engineers we got nods of agreement as they had largely reached the same conclusions. All of the presentations from the conference are available here, lots of useful insight.