If you spend much time on social media, you’ll have surely noticed the surge in AI-generated video filling up your feed.
We were promised superintelligence but instead got “AI slop” – clips-for-clicks of people and pets doing funny, diverting, but ultimately pointless things.
Or worse.
Deepfake videos of politicians or celebrities have improved in line with the advancing sophistication of AI video-generating models.
A surge of supposedly “educational” AI-generated video on platforms like YouTube Kids, say campaigners, threatens to misinform, or at best confuse, young minds.
Then there’s the film and TV industry, understandably furious that their stories, actors and characters appear to be being used to train, without permission or remuneration, AI video models.
Models so powerful they can spit out an algorithmic pastiche of their work at a fraction of the cost and effort, threatening to upend their industry.
But are we starting to see watchable, arguably worthwhile, AI-generated content out there?
It was a young, blonde, American Instagram influencer that stood out to me. No different to any other of her TikTok generation, except for the fact that she can travel through time.
Chloe vs History – and you’ll have to watch our full video to find out who “Chloe” really is – strikes a new balance between entertainment and education.
While such accounts might be at the least controversial end of the AI video debate, the new model behind them isn’t.
Seedance 2.0 is uncannily good.
Developed by TikTok‘s Chinese parent company ByteDance, it may have the advantage of having been trained on billions of TikTok users’ posts.
But as with all large AIs, it’s clearly been trained on video its developers don’t own.
US film studios are understandably afraid that AI, and a Chinese one at that, is profiting from them – then coming for their profits.
Though perhaps human creativity isn’t dead.
In making this report, we learned AI isn’t very good at making watchable video on its own. A disruptive tool, most definitely. A destructive one? Not so sure.
