Training Driving AI at 50,000× Real Time
General Motors, otonom araçların nadir ve beklenmedik durumlarını (long tail) çözmek için 50,000 kat hızında simülasyon ortamında yapay zeka eğitimi yapıyor. Büyük ölçekli simülasyon, pekiştirmeli öğrenme ve temel model kullanarak güvenli ve güvenilir otonom sürüş sistemleri geliştiriyor.
General Motors is developing advanced AI systems to address the complex challenges of autonomous driving, particularly focusing on rare and unpredictable "long tail" scenarios. These situations, ranging from unusual road hazards to nuanced human interactions like interpreting traffic controller gestures, are critical for ensuring the safety and reliability of self-driving vehicles.
GM is employing a combination of large-scale simulation, reinforcement learning, and foundation models to train its AI at an accelerated pace. The company is also leveraging Vision Language Action (VLA) models, which enhance image recognition with an understanding of vehicle trajectories and object detection, to help autonomous systems navigate these complex driving conditions.
Solving the "long tail" of driving scenarios is essential for the safe and widespread deployment of fully autonomous vehicles.
📌 Kaynak
Bu özet spectrumieee kaynağından otomatik derlenmiştir. Tamamı için orijinal habere gidin.
Orijinal haberi oku →