Driver less, Data Less or direction less?
Over the weekend was very happy with google play promo and watched Sully, an excellent movie and an excellent pilot. After the movie some thoughts came about, sharing them with all of you.
- The auto-pilot is there in all aircrafts but we still need 2 pilots in the cockpit at a minimum. (Drones are good but we don't have passengers in it, size is different, ranges are different).
- Ships do have auto-pilot but requires a captain and a crew.
- Some short running trains are driverless (airport transit is a good example), but passenger and good trains are not.
- Coming to the cars..for sure the car situation is different,
- We have many cars on the road with very different driving styles (My wife would agree with me on this for sure) and different rules and road conditions, which means we can collect more data in very short amount of time.
- We have unlimited processing and simulation capabilities with current data and cloud platforms, hence we can quickly correct and put them into testing (thinking DevOps) also there are less regulations I think compared to airlines, ships etc.. but do we have all the data we need?
- Supervised learning, anomaly detection, cognitive have advanced significantly for us, to now use the data effectively for machine learning.
- Finally and in my view the most important factor is the liability, in case of cars it is limited and very small compared to the planes and ships. Lawyers might have a different view though.
But.. Question and Thoughts:
- But are we collecting data from many more cars which are driven by humans and auto-pilots today, or as it seems, we are collecting data in limited fashion only in some cities and some drivers. This might not be the right data to model the driverless car on.
- Are we collecting the right set of data, don't we also require the road condition, weather condition, visibility, Driver's eye sight, alcohol levels, emotions etc. as well, across a broader set of driving styles or are we collecting and correcting the auto pilot that has already been designed? In my view we do need these to be collected unless we have all cars on the road as driverless.
- Should we not try making the trains 100% driver less as the route is fixed, the variations are less?
- Are we taking shortcuts just because the liability is quite small?
Don't get me wrong, I am a huge fan of AI, but I think there are better workloads in comparison of having a driverless car.
I do however like some areas to be human less for cars, I would rather have a no dealer interaction in the car buying process than having a driverless car. I think fully autonomous cars are still 5 years away, thougts?