During Google’s Stadia announcement at the 2019 Game Developer’s conference, CEO Sundar Pichai was discussing their progress in AI. Pichai explained how games are helping to tackle real-world challenges.
Pichai went on to explain that Alphabet-owned Waymo uses AI to run simulations in gaming-like environments. These simulations are being used to inform their autonomous vehicle efforts and the company has amassed an impressive 5 billion hours in edge case scenarios.
These edge cases are the really tricky 1% or 0.1% or even 0.001% of possible interactions that could occur when you vehicles moving through complex environments.
Using massive computational power required to run these simulations digitally, before implementing them on cars in the real world is one approach. While Waymo have some actual cars, it’s clear their emphasis on safe simulations digitally, is the complete opposite to Tesla’s strategy.
In November last year (4 months ago), Tesla achieved a milestone of 1 billion miles driven on Autopilot. By the end of this year, it’s expected that number will be close to 2,3 billion miles, thanks to the increased volume of Model 3s hitting the road.
The big difference is that Tesla is deploying Autopilot into real customer vehicles and taking the data from events where the humans have taken over, therefore identifying an edge case that wasn’t already accommodated in the system.
As this feedback comes back to Tesla, they have a to-do-list, in which they need to understand and solve for these edge cases. Naturally the scale of the challenge (and data) is beyond what humans can deal with, so both companies are using AI in the form of deep neural networks. This AI will hopefully find the answers to these edge cases and when proven would be implemented into the next OTA software update on Teslas.
It’s interesting that we’re talking about two technology companies when it comes to vehicles, not automotive companies. There’s a lot to play out to see who’s method will achieve full autonomy the fastest (my guess is Tesla) but I’d never count out someone like Google with their vast resources and engineering talent.
