An AI by any other name

Although there are many AI programs available today, they don’t all work the same way. Most large language model “chatbots” generate text by taking input tokens and predicting the next token in the sequence. However, image generators like Stable Diffusion use a different approach. Not surprisingly, the method is called diffusion. How does it work? (Nathan Barry) wants to show you that you can try it yourself using a tiny demo called Tiny-Diffusion. It creates – so to speak – Shakespeare.

With Stable Diffusion, training begins with an image and an associated prompt. Then the training system keeps adding noise and learns how the image gradually degenerates into noise. At the time of generation, the model starts with noise and reverses the process to form an image. This is somewhat simplified, but since something like Stable Diffusion processes millions of pixels and huge data sets, it can be difficult to train and visualize its operation.

The beauty of Tiny Diffusion is…