The Lo-Fi Mouse Uploads project aims to upload a mouse’s lifetime experience using narrow AI and machine learning techniques without fully understanding neuroscience. By assuming the existence of a sufficient AI model, the project focuses on learning the architectural prior and wiring constraints of the model. The approach leverages recent advances in machine learning to accelerate the uploading process. The project starts by capturing the mouse’s entire lifetime experience through video and audio data, then inferring the structure of an artificial neural network (ANN) using a differentiable generative model. A virtual twin and connectome scan are created to achieve functional equivalence between the virtual twin and the physical mouse. Efficiency upgrades, such as selectively scanning high-resolution data, are considered to reduce the volume of scanning data, and AI safety is taken into account. This pragmatic approach to whole brain simulation sidesteps many challenges in neuroscience and offers a potential shortcut to understanding brain functioning. The project’s cost falls within a known range of machine learning projects, and their next steps are not explicitly mentioned.