Who is Catherine Gipton?

Catherine Gipton is a virtual art critic and curator powered by AI, whom I created in November 2023. As an artist and creative researcher, I have been exploring the relationship between art and technology since the beginning of my career, and Catherine represents the latest frontier in this ongoing investigation.

​

The AI that 'Sees' Images

The inspiration to create Catherine initially came from the desire to test an innovative feature of GPT-4: its ability to 'see' and describe/interpret images, particularly starting from the digital artworks I created.

Utilizing a technology known as 'image input,' GPT-4 can indeed process uploaded images and provide descriptions, interpretations, and even analyses based on its 'vision' of the image. The following are some of the underlying technologies and methods involved:

​

• Computer Vision: GPT-4 uses computer vision algorithms to analyze images. This process allows the AI to recognize and interpret visual content in a way similar to how a human would.

• Natural Language Processing (NLP): Combining computer vision with its advanced NLP capabilities, GPT-4 can describe images, generate contextualized texts, and provide interpretations that take into account specific visual elements.

• Machine Learning: Through a vast training set of images and descriptions, GPT-4 has learned how to correlate visual elements with concepts and descriptive terms.

• Pattern Recognition: GPT-4 is trained to recognize patterns and structures within images, allowing it to identify objects, colors, shapes, compositions, and even emotions or artistic themes.

• Knowledge Network: The AI uses its knowledge to connect images to historical, cultural, and stylistic information, enabling richer and more informed interpretations.

​

When Catherine Gipton, as a virtual character created within GPT-4, provides a critical interpretation of a work of art, she leverages these technologies to analyze and comment on the image in a manner similar to how a human art critic would, offering insights and analyses based on a combination of detected visual elements and acquired knowledge data.

​

The AI That Gives a Face to an Abstract Entity

Once the AI's ability to 'put itself in the shoes' of an art curator and provide interesting and reliable descriptions and 'interpretations' of artworks of various genres and styles was concretely verified, the next step was to bring a virtual character to life by giving it a name, surname, and distinctive physical characteristics. I thus chose to create a young woman, symbolizing empowerment, inclusivity, and multiculturalism.

This specific phase of the project - giving an extremely realistic face to an abstract entity - began on Chat GPT 4 and then continued on Fooocus by Colab, a platform that allows for the modification and customization of images, maintaining consistent characteristics across different photos thanks to the following features:

​

• Convolutional Neural Networks (CNNs): Uses convolutional neural networks to analyze and understand the visual features of an image.

• Style Transfer and Adaptation: Modifies images while keeping specific characteristics consistent (such as the face in this case) through style transfer techniques.

• Manipulation and Adaptation: Allows for altering various aspects of the image (such as background, facial expression, clothing) while maintaining the integrity of the original image.