The SORA blockchain has been operating AI inference models specialized for drive state monitoring using CPU double-precision floating-point operations.

Now, we are significantly expanding this with the blockchain, introducing GPU single-precision floating-point operations compatible with cutting-edge large language model (LLM) inference, while also supporting traditional AI inference models.

Through the already implemented quantum-resistant NFTs, we offer an innovative approach that enables GPU single- or double-precision operations to be exchanged via the blockchain on SORA. This achievement will not only support mining but also provide inference services, creating a new business opportunity.

The following images and videos are powered by SORA’s AI.

Among NVIDIA GPUs, the GeForce series is specialized in single-precision performance. Even the latest models don’t offer significant double-precision capabilities, so we primarily utilize GeForce GPUs that can output decent single-precision performance. However, for SORA’s AI, which focuses on physical simulations like drive control, double-precision is essential. In that case, from a cost perspective, AMD’s RADEON becomes a viable alternative to NVIDIA.

The single-precision GPU is expected to be utilized for purposes other than drive-related tasks. As outlined in the roadmap, I want to experiment with a method that improves security by having the system learn from transaction patterns. Although the impact may be slight, I will build this approach with the perspective that, if it can significantly enhance coin protection, it would be worth pursuing.