Microsoft and MIT are working together on a project to eliminate the so-called virtual “blind zones”, which lead to errors in machines that are managed without human intervention.
Roads are extremely unpredictable places – especially when real people are driving. Preparing a car for independent driving in any possible situation is a monumental task.
The AI, developed by Microsoft and MIT, compares the actions taken by people in this scenario with what the AI will do without a driver. Where the human solution is more optimal, the behavior of the vehicle is updated.
Ramya Ramakrishnan, the author of the report, said:“The model helps autonomous systems to better understand real situations on the roadway. Often, trained simulations are not able to soberly assess the real situation make mistakes – for example, getting into an accident.The idea is to use the human driver’s intuition to safely bridge the gap between the simulation and the real world so that we can minimize the likelihood of an accident.”
The researchers made sure that the AI without a driver did not consider its action to be 100% safe. Using the David-Schen machine learning algorithm, AI uses probability calculations to determine patterns and determine whether an action is really safe.
We have not yet reached the stage when the technology can be actively implemented. So far, we are limited to testing video games. Nevertheless, the results of recent studies give great hope that an AI without a driver will one day be able to safely respond to any traffic situation.
Editor of IMD News