How graphics cards became the heart of artificial intelligence

Graphics cards, originally designed for image rendering, have become one of the most important technologies of the 21st century. Their key advantage was the ability to perform parallel processing.

Over time it turned out that parallelism is invaluable far beyond gaming — in scientific computation, physical modeling and big data analysis. This opened the door to AI applications.

Traditional CPUs were designed for sequential processing. GPUs, with thousands of cores, handle the massively repeated mathematical operations required by machine learning far better.