CMU Releases New Tool for Experimenting with Text-to-Image Generative AI


Carnegie Mellon University graduate student Adithya Kameswara Rao and his advisor, David Touretzky, have released a new educational tool for experimenting with text-to-image generation. The tool, called DiffusionDemo, is built on Stable Diffusion, a popular open-source model from that generates novel images given a textual description. Other models in this class include DALL-E, Imagen, and Midjourney.

DiffusionDemo uses a scaled-down version of Stable Diffusion that trades image quality for speed, allowing rapid experimentation with different prompts and parameter settings. The demo includes ten panels that support different experiments or views of the image generation process. The key educational goal is to help people understand the notion of "latent space," a mathematical construct in which images are described. Nearby points in latent space describe similar images. A neural network called a "decoder" maps these descriptions into actual images.

DiffusionDemo can be accessed at

DiffusionDemo was developed with funding from NEOM Company and the AI-CARING AI Institute (NSF award IIS-2112633).

CMU Introduces Powerful Educational Tool for AI Understanding in Middle Schools

Neuron Sandbox

Carnegie Mellon University (CMU) researchers have launched an educational tool, Neuron Sandbox, with the aim of equipping middle school students with an understanding of a foundational element in contemporary artificial intelligence. Neural network technology serves as the foundation for a wide range of AI applications, including convolutional networks used in computer vision and transformer networks seen in large language models like ChatGPT.

As an interactive tool, Neuron Sandbox guides students to understand a common type of simulated neuron called a linear threshold unit.  It helps student visualize both the neuron and the truth table that describes its inputs and outputs.  It presents a series of problems of increasing difficulty which students can solve by adjusting the weights on the neuron's input connections and/or its firing threshold.

Neuron Sandbox is available now and runs on any web browser at:!

Neuron Sandbox is designed to be used by middle school students, but it contains advanced features that make it useful for high school and undergraduate students as well.  In addition, the tool includes an interactive editor that allows teachers to create their own custom problems for their students to solve.  

Neuron Sandbox was created by CMU students Angela Chen and Neel Pawar, along with their advisor, Professor David Touretzky. Touretzky is also the principal investigator on an NSF ITEST project aimed at developing a 9-week artificial intelligence elective for middle school students in Georgia and a member of the AI-CARING team. "We're excited to test Neuron Sandbox with our middle school teachers and observe how their students respond to it," Touretzky stated. "We're continually adding new features to the tool, such as the ability to switch between a threshold display and a bias connection display, to enhance its usefulness for more advanced students as well."