With the development of technology, the automobile industry is going through radical changes. Currently, the two main emerging areas in this industry are electric cars and autonomous driving vehicles. Although autonomous driving has come a long way recently, it has some shortcomings. A researcher named Mehul Bhatt explains that she has found solutions for these shortcomings by working with her partners in Germany and India.
Unlike artificial intelligence that provides autonomous driving in traffic, people always keep their perceptions open about what may happen unexpectedly. Artificial intelligence technologies, on the other hand, act ‘according to existing information’. This poses some problems that need to be overcome. The study, published in the journal Artificial Intelligence, argues that a new technology created by combining modern “deep learning” methods with the perception of “logic” will bring autonomous driving on a par with humans.
The new artificial intelligence will be able to explain the reasons for its decisions:
Mehul Bhatt, who states that only artificial intelligence technologies based on ‘machine learning’ lack human-specific capabilities, explains the differences of the new artificial intelligence technology planned to be developed as follows: For example: a cyclist goes out of sight of the car when he gets behind another vehicle. The new technology will be able to understand that the cyclist is still there and will become visible again. Such human abilities are an element lacking in artificial intelligence in general. ”
Stating that this kind of ‘logic framework’ is essential in autonomous driving technology, Bhatt also draws attention to reliability. Pointing out that the current artificial intelligence does not explain the decisions they make in traffic, Bhatt states that this problem will also be overcome. With this new method, artificial intelligence will be able to explain the reasons for the decisions they make in traffic. Bhatt says that this way, we can get clearer results about the correctness of the decisions taken.
39585
