On 5/2/2024 11:32 AM, Peri Hartman via EV wrote:
The question i have is, how does the "decider" know when the driver is correct
versus the algorithm ? In others words, what keeps it from learning human error ?
One of the things they do is watch for "interventions". i.e. the User
suddenly takes over. That indicates something went wrong. They add
that case to the training, as well as variations of it - since they can
now also create simulated data to train the AI with.
It also gets trained with a lot of "just normal driving" data.
One of the interesting things that happened when they switched to an all
AI setup, is that the AI figured out what a Stop Sign meant without that
having to be explicitly forced in the training.
_______________________________________________
Address messages to ev@lists.evdl.org
No other addresses in TO and CC fields
HELP: http://www.evdl.org/help/