NVIDIA offers a variety of mission-critical products that cannot tolerate the errors common to most AI-generated code. This includes GPU drivers that support everything from basic gaming to large-scale AI training and inference operations. The company likely has strict guidelines for newly generated code and requires extensive testing before deploying the code to production. This isn’t the first time NVIDIA has used AI-assisted workflows in its products. The company already has a dedicated supercomputer in place that has been continuously enhancing its DLSS (Deep Learning Super Sampling) for several years, and some chip designs have been optimized using in-house AI tools.
Not all uses of AI are necessarily bad. NVIDIA and Cursor report that because NVIDIA generates three times as much code, the bug rate remained the same despite several company-specific optimizations being implemented. Some interesting AI-assisted developments have resulted in DLSS 4 and a GPU die that is 25% smaller than comparable industry standard solutions. When used for interesting applications such as those demonstrated by NVIDIA, development can be highly beneficial to both gamers and businesses. We’ll have to wait and see what new technologies (or bugs) come out of the new software development pipeline.
