Why was CUDA created?

Key Takeaways

We found I was at Nvidia and we had all these people these scientists that were trying to build fluid dynamic systems and thermodynamics, electromagnetic simulation using GPUs and we're like, "How are you doing?" Well, they're they're programming using an API a graphics API DirectX or OpenGL. That's what's called a kludge, right? I'm taking a language and I'm translating it just to get the job done. And so that's when we said, "Wait, we should go build an a compiler. We should build a language that actually talks to the GPU in a C-based programming language and that was basically CUDA. So the customers were kind of telling us that they wanted to do high performance computing and then AI and you know, convolutional neural nets came, transformers came and you tell me what comes next. I don't think the transformer model is the end all and be all. I think it's a approximation of knowledge using extreme amounts of compute and you know, high level of pattern recognition on large amounts of data. It actually works, but it's kind of a brute force method of doing it. So I think we're going to find new ways and new models and