I’m backdating this one: five years ago (2018), I predicted that in ten years (2028) the entire content production pipeline would be automated. We seem to be well on track.

In the immediate future, AI will become an integral part of our creative process. Already you can compose simple scenes and use neural rendering to enhance them, without spending countless hours perfecting the materials and lighting1. Similarly, the painstaking process of detailing is being replaced with AI-powered inpainting and image-to-image style transfer2. And the adaptation of existing images to new formats and media is being revolutionised by outpainting3.

Currently, the creation of a finished piece requires the use of multiple tools. One might be used for backgrounds, another for characters, and yet another for high-quality faces and intricate details. The process becomes even more complex when creating a video, requiring additional tools for tasks such as frame interpolation. These AI tools still require skilled guidance to produce high-quality results. Some argue that these limitations will persist in some form and so we have nothing to worry about. But here’s my prediction:

  • 2024: Within the next year, we’ll see a new generation of tools that are better integrated and capable of creating still-image artwork that requires no manual rework. Text-to-image prompts will become more intuitive, eliminating the need for specialist knowledge of specific terms. We may even see the dawn of end-to-end (automated) creation of larger pieces, such as comics and short films.
  • 2028: Fast forward five years, and I believe we’ll be in an era of automated creation of complex works. Think TV episodes of daytime TV quality or better. Created in minutes per frame, this would put the power to create TV shows in the hands of laypeople. We’ll still be publishing mostly via platforms like YouTube, but we’ll also start to see specialised services that allow people to share generative experiences with each other.
  • 2034: In the longer term, expect a world where AI can automate the creation of feature films. Content would be hyper-personalised for the audience, allowing you to modify things you don’t like, such as long battles and annoying characters.

These predictions are based on the progression of narrow generative AI. If AGI is achieved within this timeframe, the timelines will be drastically altered. An AGI, by definition, would have the capability to accomplish all of these tasks and more.

Nick Cave wrote an excellent piece on what this will mean for human expression.

It could perhaps in time create a song that is, on the surface, indistinguishable from an original, but it will always be a replication, a kind of burlesque.4

I’d add that, economically, it doesn’t matter whether it can make genuine works: it only needs to be technically good enough to capture the vast majority of creative work. The economic and social impact will be enormous. We should be preparing for it now.

  1. Blender × Stable Diffusion plugin ↩

  2. Photoshop × Stable Diffusion integration ↩

  3. DALL·E demos on Instagram ↩

  4. “Chat GPT - What do you think”, Nick Cave, Jan 2023 ↩