How Far Will AI Video Go?
Mapping Out the Scenarios
Starting April 2025, all full posts, including archived posts, will be available on my Substack, The Mediator.
I often write that the last 10–15 years in video(1) have been defined by the disruption of content distribution and the next 10 years are poised to be defined by the disruption of content creation.
Here’s the argument: The internet unbundled information from infrastructure and, with the help of a host of related technologies and massive infrastructure investment, caused the cost to move bits around to functionally head toward zero. We know what happened next.(2) Now, there is another emerging general purpose technology, GenAI, that may send the cost to make bits to head toward zero, too.
This symmetry of falling costs to move bits and make bits sounds good. It’s pithy and memorable. It seems plausible. But still: it is admittedly very high level and hand wavy.
What will GenAI really mean in practice for the video business? Will the cost to make TV and movies truly “fall to zero?” Will two kids in a dorm room one day make the “next Avatar?” Or, is GenAI another flavor of Silicon Valley’s naïve technological determinism, a blind belief that technology always marches forward and anything that’s technically possible is inevitable, without regard to pesky inconveniences like law, regulations, ethics and consumer demand? And what does disruption mean, anyway? Are we talking about complete devastation, the Kodak-disrupted-by-digital-cameras kind of disruption, or the far more benign Marriot-disrupted-by-Airbnb kind of disruption?
Figure 1. Two “Victims” of Disruption
The only credible answer to these questions is: no one knows. That doesn’t mean we’re completely flying blind though. We can frame out a range of possible outcomes by using scenarios.
Tl;dr:
- Scenario planning is a useful tool for navigating uncertainty. It can help identify the range of possible outcomes, the key milestones to watch, and the potential implications.
- A key step is identifying the two critical variables that will determine possible future states and the extreme potential outcomes for each. Below, I use technology development and consumer acceptance to construct a scenario matrix and analyze the possible state and implications of AI video in 2030.
- The possible outcomes for technology development range, at one extreme, from AI video models stalling out at their current capabilities to, at the other, completely resolving their current limitations in realism (especially the “uncanny valley”), audio-visual sync (especially lips), understanding real-world physics, and fine-grained creative control.
- The possible outcomes for consumer acceptance range from skepticism and sometimes outright hostility to fully embracing AI (and actually preferring it for some use cases). Steps along the way include consumers accepting it for certain content genres and use cases, especially those that don’t rely on emotive humans.
- Varying each of these variables between their extremes produces a 2 x 2 with four scenarios: low tech development, low consumer acceptance (“Novelty and Niche”); high tech development, low consumer acceptance (“The Wary Consumer”); low tech development, high consumer acceptance (“Stuck in the Valley”); and high tech development, high consumer acceptance (“Hollywood Horror Show”).
- Writing out narratives for each scenario is the most instructive part, because it helps make the abstract more concrete.
- Reality will probably fall somewhere in between, but this shows why it won’t require the most radical scenarios for the video business to change radically.
Click here to continue reading the full post on my Substack, The Mediator.
