William Bartlett

Executive creative director at Framestore London

Right now we’re at the early stages of the AI revolution and a lot of the focus has been on how we use AI tools to speed up and augment our existing workflows and creative structures. This trend will certainly continue but alongside it, we are also developing entirely new ways to assemble the patterns of pixels that make up images. Midjourney, Stable Diffusion, Pika and Runway have all become part of the day-to-day tool sets that VFX companies are using.

There are many questions about a fundamental ceiling of creativity that AI will reach and if it can ever be anything more than a ‘creative partner’. Some limits have been seen with the first wave of tools but new models like Google’s Gemini [AI language model] will evolve the complexity of generation as they learn across different modalities. Up until this point, large language models (LLMs) have broadly been creating original writing based on huge datasets of language. Multimodal models will be able to genuinely build in inspiration for original writing from music, images, sound and eventually even touch. And likewise, this cross-pollination of perceptual stimulation will work in all directions, exactly like it does with our own creative output.

Alongside this, the further evolution of tools for conditioning and controlling outputs will broaden the scope for real world applications and I think we will continue to see a rapid expansion of the AI footprint in the creative arts