Opinion: The Great AI Slop

The great AI slop-ification has begun. Earlier this month, OpenAI released the Sora app in the US, which uses its new “Sora 2” video generation model to serve up short-form content. Think TikTok, but all AI-generated and even more addictive. Slop.

Nobody asked for this. We aren’t even close to sorting out the harms caused by existing social media. But now OpenAI has turned the dial up to eleven.

What does it offer? Well, you can make a 5-second video based on nearly any prompt you want. Like ChatGPT, but with video output instead. Sounds fun, right? It is, until you realise that these video creations are near indistinguishable from reality.

A video of Lorde doing a backflip? Of course. Michael Jackson stealing KFC? Coming right up. What about an AI Christopher Luxon announcing that a tsunami is about to hit Auckland? The resulting panic could be deadly.

It’s not just public figures, either. Sora invites you to upload your own identity to the app. Then, it will gladly insert you into any video you like. And your friends, if they upload their identity as well. Deepfakes will be easier than ever.

We’ve known for a little while that we can’t be sure what we read online is written by a human. Now, any video could be fake, too. The Dead Internet theory is coming true: about half the text you read online is AI-generated, with AI image and video quickly catching up.

What does this mean for our online lives? One option is that it becomes the new “normal”, where human and AI content co-exists and we learn not to trust any of it. We can’t detect it reliably, and there is no easy way to force the labelling of AI content.

Another theory says we will disengage and go back to the real world, reminded of the value of human relationships. Though robotics is also advancing fast, so we may see AI with a physical presence before too long.

AI video generation comes with other costs. A recent study showed that a single 5-second video used the same amount of energy as running a microwave non-stop for an hour.

Due to the way the AI works, the longer the video gets, the energy use ramps up more and more quickly. A 10-second video would be equivalent to the microwave running for four hours. In contrast, a single response from ChatGPT is only one-tenth of a second of microwave use.

Optimists will say that the technology will get more efficient and the energy use will reduce quickly. That’s true. But that will just allow even more widespread use of the tech.

OpenAI is aiming for 250 gigawatts of energy production by 2033: more than the whole country of India and around 25 times that of New Zealand. And OpenAI is just one company.

We have to be optimistic too. Otherwise, Big Tech wins by default. But what can we do about it? Alongside 20 other AI experts, I’ve co-authored an Open Letter calling on our politicians to show some leadership and commit to risk-based regulation of AI in New Zealand. Over 500 more people have signed it since. You can, too, at .

Big Tech is haemorrhaging billions of dollars and needs us to use their products if they are to survive once investors lose patience. There are already cracks appearing. Nearly every AI video I see on Facebook has screeds of anti-AI comments. My 1st year students dislike it even more than I do. And everyone is talking about the “AI Bubble” now: these companies are so over-valued, it’s near-impossible to see how they could ever earn enough returns to justify their stock prices.

I became an AI academic because I could see the amazing opportunities that the technology could bring. AI is bringing us personalised healthcare, advanced weather forecasting, and so many other benefits. Why can’t we spend billions on these instead?

We, as a society, must reject AI slop and all the harm it causes. We have a choice, and we must put ourselves first.