Self Driving Is All or Nothing

April 6, 2024

The graveyard of self-driving car projects is full, and for good reason. I think the technology must reach near-100% perfection before it can become a part of our day-to-day reality.

Consider Amazon’s recently cancelled "Just Walk Out" technology. Pull back the magic AI curtain, and you’ll find 1,000 workers in India manually verifying shopper’s transactions.

I think their AI technology was stellar. It was probably 99.99% perfect. Yet they had to rely on human labor, because incorrectly billing 0.01% of shoppers is unacceptable. And relying on human labor, well uh... defeats the whole purpose of the technology.

My guess? I think Amazon cancelled the project because 100% was not attainable, at least in the near-term.

I believe the same could be true for self-driving: it’s an all-or-nothing problem. If a company’s self-driving system is 99.99% perfect, it’s halfway there. The last 0.01% is what makes it an extraordinarily challenging problem.

This is my (somewhat uninformed) opinion of course. I still believe it to be a worthwhile problem. I also believe near-100% is attainable, in time. In my time, though? I’m not sure.