The company Open AI has presented the new service Sora. In a press release, 48 different films are shown, which quickly gained wide spread and attention on social media.
One of the films shows an American 19th century society from the air. Along a road, people make their way on foot and on horseback – but the images are not from the 19th century, they are not even real.
The films are generated with the help of AI and the video is created based on the sentence “Historical images of California during the gold rush”.
The new service Sora is described as a text-to-video tool and should be able to create one-minute films based on a short so-called prompt, an instruction that describes what the tool should do. Open AI has previously offered its users to embarrass images, but not videos.
Currently, the service is not available to the public. Developers conduct risk assessment while designers and filmmakers evaluate it.
There are limitations and mistakes
Although Sora has been praised by many for its realistic films, Open AI believes that there are some problems with the service. Even Martin Berg, professor of media technology and associate professor of sociology at Malmö University, sees some limitations with the tool, as it is trained based on a certain type of visual data.
– What happens if you write in that you want to see a really cool autumn day in Malmö? With horizontal rain and the depressing despondency right before payday that only this town can produce. I don’t think it would clear it up, says Martin Berg.
Despite the fact that the published films impressed many and were widely distributed, Martin Berg believes that there are things that make them distinguishable from real images.
– There are many details. Like this woman who seems to be walking around somewhere in Tokyo, where it has just rained. But she is completely dry and also moves in the middle of the street. There is no sign of foul weather on her.
In the films that were presented, only images have been generated and the press release does not mention anything about future sound.
– I think you can miss sound in this. It’s just the visuals and then you wonder what would happen if we take this Japanese series. Will it capture Japanese sounds and the kind of acoustics created when it has rained outside or when it is raining? There are many such things, says Martin Berg.
Difficult boundary around what is real
The warning bells for the development of AI have been many and there is a certain fear as the technology advances rapidly.
– One reason why people might think this is unpleasant is that we suddenly get the opportunity to put pictures of our thoughts, says Martin Berg.
Even if developments make it harder to distinguish between reality and AI, Martin Berg believes that people’s mindsets are evolving – we are becoming more skeptical.
– It leads to a difficult line between what is true and what is false or misinformation in various ways, I absolutely believe that. But perhaps it is also the case that we are moving towards a situation where sooner or later we distrust everything. Then one can rather ask the question what kind of almost postmodern hyper-reality we live in. Where everything is somehow half-faked.
He also points out that it is important to talk about how technology can contribute to potential disinformation campaigns.
– With deepfakes and the like, but we already have that today, says Martin Berg.
At the same time, not everything is negative and there are some quick concrete things that can be improved. According to Martin Berg, the new technology can give us a chance to visualize and put our thoughts into words.
– Imagine if this type of technology could help us visualize what is difficult to put into words. Maybe we can communicate our dreams for once. At least it might be more fun Powerpoints in boring workplaces, says Martin Berg.