What is the endpoint of AI content creation?
We have all seen firsthand how quickly AI content creation has developed, in text, images, and video! This comes the caveat that the copy produced must always be reviewed and edited by an all-important human writer, but it did get me thinking about what the endpoint of this technology could be, or whether it even exists.
Paying for the trainer, not the trained
As AI copywriting continues to improve, those training models would quickly become a commodity. It’s perfectly feasible that a customer ‘rents’ an AI model that has been trained by a specific writer, essentially bringing the style of a well-known author into the room. It doesn’t matter what the planned output would be, whether it is a novel, company communication, or simply website wording. The value in each model would be placed firmly on the trainer, with higher price points set for well know names. Think of it like bringing in your favourite crime writer to add some exciting turns of phrase to your company blogs.
As with all AI innovations, there is a question around responsibility and where it lies. If I purchase a trained AI model, use it to create content based on my own prompts, and am not happy with the results, whose responsibility is it to provide a fix? Was it my prompts? Was it using the style associated with unfamiliar media? Would the original author and AI trainer have to provide user guides to get the best possible results? While all these questions are hypothetical, they show that there will always be work for people to adapt AI-produced content to fit the desired purpose.
On the B2C side, the question is much broader. As a consumer, would you pay the same for a novel that was produced completely by AI but in a model trained by your favourite author? … perhaps you have already.
Indistinguishable video content
In the early days of technology it was very easy to spot AI-produced video through tell tale signs like too many fingers or unnatural movement. While it is getting harder to spot these signs every day, they can still be spotted by viewers in the know.
But what about the next stage? Similar to content creation, what happens when an AI-produced video is indistinguishable from real life footage? The first concern is obviously trust, if any video could be faked then how would we ever know what was real? CCTV footage or video evidence could be called into question for legal processes, but luckily we are a fair way away from photo-realistic AI video generation.
On the consumer side, the technology journey could be very similar to written content, in that AI models trained by certain directors could be leased out for the viewer. Imagine heading home after work and instead of deciding what movie to watch that evening, you decide what movie to make that evening.
In the ultimate fan fiction, you could bring together your favourite actors to appear in a sequel that was never made, creating scenes either as a. Again, in this scenario the question of the responsibility is up for debate. Who is responsible for the quality of the output? Would actors be able to turn down an AI-prompt as if it was a part they didn’t want to play, or could they earn a licence fee for use of their image in such models?
