To achieve autonomy on the motorway, we need to modify the motorway as well as the vehicle.
When you use “artificial intelligence” instead of the more accurate “probability modelling”, you create a bad legacy.
This isn’t the first time people have inflated expectations from “artificial intelligence”. It happened again 50 years ago.
The driver is not just someone who monitors the road ahead and decides when to press the brakes. The driver is the person responsible for the vehicle while it is on road.
What will a driverless truck do if the engine is making a strange noise?
Computer vision and other modern technologies are great and can be used to achieve autonomy in controlled environments.
Recently there have been several experiments where a truck goes out driving “on its own”. Except that there is actually a driver in the cabin.
If Atlas looks so impressive, it’s because it looks like us. It’s not because it has any intelligence at all.
The Altas robot is impressive, but it is entirely unintelligent.
Our progress in vehicular automation doesn’t mean we are necessarily on the right path to level 5 autonomy.
Level 5 autonomous driving is not just a step beyond level 4—it’s a whole different world.
Trying to foresee all possibilities won’t solve the problem of true vehicle autonomy.
A truly autonomous vehicle must be able to handle this situation.
To take the driver out of the cabin, there’s only so much we can do on the vehicle; we also have to control the road.
Level 5 autonomy cannot be achieved by current technology.
“Artificial intelligence” is just a fancy term for “probability modelling”.
The terms “driverless” and “autonomous” are misleading.