Unconventional AI Inc. has developed an artificial intelligence architecture that could improve the power efficiency of image generation models, forming the basis of a new neural network series, Un-1, released on Thursday. The company is led by Chief Executive Naveen Rao, the former corporate vice president of Intel’s AI platforms group. In December, the company raised $475 million from a consortium that included Amazon.com Inc. founder Jeff Bezos. The Un-0 model series is part of an effort to develop more efficient AI chip architectures, optimized to run on oscillators rather than standard transistor-based circuits. Un-0 generates images using several thousand simulated oscillators, which are linked together so signals from one virtual device affect the output of others. There are six Un-0 models varying in size and output quality; the smallest comprises 1,024 virtual oscillators while the largest features 16,384. Unconventional AI trained the models using two open-source datasets, CIFAR-10 and ImageNet-64, that contain thousands of images optimized for machine learning projects. The training process involved calibrating how Un-0’s simulated oscillators affect one another and the frequency at which they generate signals, rather than optimizing weights.