OpenAI achieves the next wow! with Sora

February 19, 2024

Hardly any of us would have thought that OpenAI could do it again: Sora gave us the same aha moments as ChatGPT did just over a year ago. Wow - we didn't expect that: That AI-generated videos of this quality would be possible in February 2024 is surprising. Many had predicted that something would happen with videos. But in February and with this quality?

OpenAI has published about forty videos on the Sora website that demonstrate the power of its new language model for video generation. Everything is represented, from short animated films and nature shots to complex scenarios with moving people. The language model is still under development, both in terms of security-related issues and video quality. It is not yet known when we will be able to prompt with Sora ourselves.

The videos are a maximum of 60 seconds long and still have some weaknesses. However, compared to the AI-generated video footage of the past year, the two left paws of a cat or the strange reproduction of the wolf pups are negligible. Especially as we have seen a rapid evolution in images over the past year from deformed hands with six fingers and humans with three legs to photorealistic examples that are indistinguishable from real photographs.

What does Sora mean for us?

At the moment, the focus is rightly on the social impact. In his column, Sascha Lobo coined the term "fake reality", which aptly describes it. We can no more be sure of the medium of moving images than we can be sure of text and images.

But beyond the dystopian prospects of perfect propaganda - what else can we expect?

In a business context, Sora is first and foremost a stark reminder that we cannot rely on our business models. For those who may have become somewhat immune to last year's predictions about how much generative AI will impact our work over the past year, Sora is a second wake-up call. We are faced with the task of redefining where the strengths of our own intelligence lie and what human labor will accomplish. A simple "AI cannot replace us, we need humans in tandem with AI" is not enough.

Read more