When the developers of artificial intelligence for the
autopilot are convinced that he has learned to perform regular tasks, they are
faced with the task of testing their offspring in obviously dangerous
situations. But where to get so many unpredictable drivers, how to create
unforeseen situations on the road and involve AI in resolving conflict issues,
not injuring anyone or anything? Intel came up with how to adapt for this
purpose the most infamous game, where permanent chaos on the road is the norm.
It's convenient - those who, even though sometimes just
drifting around the virtual world of Grand Theft Auto, without chases and
skirmishes, certainly paid attention to the high detail of the aspects of
traffic. There are signs, traffic lights, pedestrians of varying degrees of
impudence and both obedient and inadequate chauffeurs. And they sit behind the
wheel of a wide variety of vehicles, so that the required variety of traffic is
provided.
Before a drone does not need to go through this or that
adventurous mission, on the contrary, they learn to drive safely in a
constantly changing environment. This is the appearance of animals on the road,
and the weather change, and contact with the police, and travel through very
complex interchanges and intersections, as well as training in building the
optimal route.
An additional argument in favor of learning autopilot on the
streets of the game world - engineers can simulate any situation and run the
test countless times. Up to drive 3 million km per day, meticulously marking
how the AI reacts to minor changes in the scenario. However, there are doubts
that the supervisory authorities will count these virtual trips as a real
experience in training autopilot systems. Therefore, while everything is
theoretical, and the "robbed GTA" robot will not soon leave on a real
road.
Comments
Post a Comment