Autonomous Car - so kool !

Who will not like to see Autonomous, Intelligent car in a movie? And if it becomes a reality? won't it be so cool, right? It's a dream, figment of imagination and amazement of a child. And a project of amusement for many. I wonder, what is the need and business case behind spending so much to achieve a fully automated car. Is it to create a sequel of "Total Recall" without using animation or to achieve a personal goal like "blue moon". 



Safety
Tesla, Waymo and many have stated safety as the objective behind AV project. Well, 97% of road accidents are attributed to human error! So humans have to be replaced with AI. But why don't we spend in human? In the end you can get same result by spending in either Tech or Human. 


Then why are tech companies spending so much in Technology? Because spending in Y axis makes them "high" and "kool"; that is why the axis is titled "K".

Making Money

In Elon Musk's words "Enable your car to make money for you when you aren’t using it". Conventional vehicles sit idle 95% of the time. Mobility-as-a-Service (MaaS) is the future. Uber, Ola etc are old school and not exactly MaaS as a consumer still has to see a human face and transact with it. Won't it be kool to place an order from a phone, the driver less car reaches you, takes you to destination and you pay on phone. No human is involved and no error is made. So "Kool". 


More serious talk and less satire

There are 5 levels of autonomous driving: Level 0 to Level 4. I strongly believe, like many normal curves, these also will follow a normal curve in giving value to human kind. 


At L2, the AI is complementing drivers and can give the best safety result. L3 requires driver during critical situation. Don Norman wrote in chapter "The Paradox of Automation" : "When automation works, it is wonderful, but when it fails, the resulting impact is usually unexpected and, as a result, dangerous...". It is foolish to expect human to respond quickly in a critical situation when it is relaxing in the autonomous machine. Even in a normal driving situation, human takes about 1 sec i.e. 20 mtr in distance to respond when it is fully attentive.  

In one mckinsey article, it rightly points out: training a car for 95% of situation is still easier. But teaching a neural network up to 99% correctness is a difficult and long journey. "If you take a regular human driver, on average, takes roughly 165,000 miles to meet an accident. That means, in 99 percent of the cases, the human person is right when they drive themselves."


Then why are techs investing and marketing so much on AV?

There is an adage from an experienced person (me) "you prepare for IIT to get admitted to NIT" (IIT is one of the toughest exam in India for Engineering stream and NIT is the 2nd best option). And it is "Kool". 



Comments

Popular posts from this blog

Are EVs eco-friendly in India?

Field force tracking app - free app made using Google services in weekend !

COVID: LoC and my need of premise to blame