Grappling with Technical and Ethical Challenges Facing Self-Driving Cars at AutoSens

Last week, hundreds of engineers working with automotive sensing technology attended AutoSens Detroit to be among colleagues who, like them, are solving difficult challenges in vehicle perception. Several folks from Mighty AI attended, including our principal computer vision engineer Bernd Heisele who presented on research he recently conducted regarding the accuracy of deep neural networks for vehicle detection. (Spoiler alert: human annotators are still far more accurate than the most popular open source vehicle detection models available today.)


While there were dozens of compelling topics and themes discussed throughout the event, perhaps the one that bubbled up most frequently was: how will we know when these vehicles are safe?


Consistent throughout the deeply technical presentations were plenty of conversations about why it all matters and why engineers are driven (pun intended) to do this work: the chance to save lives. It’s clear that people working in this industry at all levels are grappling with safety and regulations, wondering how it will play out. In the U.S. for instance, as of 2017, only 33 states had introduced some type of autonomous vehicle legislation. With no broadly-applied set of local, state, federal, or global guidelines in place for autonomous vehicle legislation or safety certifications, as of today, the industry is largely relying on self-certification.


At Mighty AI, we think we will soon start to see regulations for self-driving cars that require certain amounts of testing and validation for new systems. We believe there’s a critical need for the industry to work hand-in-hand with regulators to develop those guidelines. The technology is accelerating at such a pace where all parties must be in lock-step to ensure regulations are effective in keeping consumers safe.


This includes going beyond a series of “check all the boxes when testing” requirements to drilling into, “did you follow a reasonable engineering process?” On the data side, it requires taking a systems—vs components—perspective of your data. When developing autonomous vehicle systems, engineers must ask themselves, “how will this data impact the whole system, beyond just the one component I’m developing?”


AutoSens Detroit 2018 was a great opportunity to dig into meaty technical and ethical questions facing our industry. We look forward to seeing both familiar faces and new ones when we head to Brussels this fall for the European 2018 edition of AutoSens!

Source link

Related posts

Self-organization in aggregating robot swarms: A DW-KNN topological approach.


[SSCAIT] StarCraft Artificial Intelligence Tournament Live Stream


AI used to rank NHS patients in order of urgency to clear COVID-19 backlog


This website uses cookies to improve your experience. We'll assume you're ok with this, but you can opt-out if you wish. Accept Read More

Privacy & Cookies Policy


COVID-19 (Coronavirus) is a new illness that is having a major effect on all businesses globally LIVE COVID-19 STATISTICS FOR World