There’s a story from the early days of cinema that appears to be at work in Sora, a text-to-video creation tool that OpenAI launched this week. And given that Sora’s servers are suffering from demand, with many OpenAI subscribers still waiting to try it out, we have news time.
You probably know Train Arrival at La Ciotat station (1896) by the Lumiere brothers, even if you’ve never seen it. Like Sora, the Lumieres created very short movies that showcased the latest technology. We are talking about the cinematograph rather than the AI ​​rendering, and a fancy 50 second movie rather than the maximum 20 seconds allowed in Sora videos.
Still, it’s the same principle: This was an early look at a new form of horror entertainment. According to the legend – the legend that was inspired by Martin Scorcese’s charming film about a boy from the Lumiere era, Hugo (2011) – The Arrival of the Train the audience ran in panic from the engine that seemed to be headed straight for them.
A similar sense of panic clings to Sora – specifically, panic about what AI videos can do to further disrupt our “post-truth” media landscape. The average viewer already has a hard time judging what is real and what is real, and the problem gets worse when they are stressed. We live in a prime age of conspiracy theories. The world’s richest man has already shared an AI-powered video to help sway the election.
Can an AI chatbot talk about believing a conspiracy theory?
What happens when Sora does anywhere an instant look as real as something you’d see on the evening news – ready to be shared on social media?
OpenAI seems to think that its watermarks, visible and invisible, can prevent any shenanigans. But having downloaded dozens of Sora videos now, I can attest that the visual watermark is small, illegible, and often fades into the background. It would be child’s play for video editing software to cut it perfectly.
So a world full of misinformation on purpose, from bad political actors or influencers trying to end their dealings, comes down on us like a train. OK?
It’s wrong. Because as the true story of the Lumiere film tells us, people are actually much smarter about new video entertainment than we give them credit for.
Here’s the thing about it The Arrival of the Train: the legend is probably wrong. We have no evidence that the audience ran away from the cinema, or even panicked when they saw the train coming in the 50 second clip.
Media studies professor Martin Loiperdinger calls the horror story “cinema’s founding myth,” and notes that it can be traced back to literature written in the second half of the 20th century. It is possible that the writers combined it with an experimental 3-D version of the Lumieres later The Arrival of the Trainwhich screened several times in 1934 and was like many 3-D movies to come – a novelty, and a commercial failure.
So no, early audiences probably didn’t confuse a motion picture train with a real train. Instead, they seem to agree with the whole idea of ​​the movies very quickly. Current accounts of Lumiere shorts (of which there are many; The Arrival of the Train was not seen as a stand-out) are full of excitement about the opportunities that are now open.
Mashable Light Speed
“Why, if this continues,” wrote another newspaper, The Courier de Parisin 1896, “we can almost overcome the loss of memory, we can almost do away with separation, we can almost do away with death itself.” (Spoiler alert: we didn’t, although that sounds like a good 19th-century premise. Black Mirror episode.)
Another magazine, La Science Francaisdelighted in the “wonderful sorcery” that created the “hallucinatory phantasmagoria” of the cinematograph. Even today’s most enthusiastic AI boosters would have a hard time agreeing to Sora on the same terms.
Because like most AI, Sora is often “hallucinatory” – and not in a good way.
7 Sora videos blew up social media after the launch
As I found out from the times that the OpenAI servers were not hit, almost all of the videos produced by Sora have some details that look wrong to the human eye. I typed the command “reporter’s desk is crashing due to lack of access to AI videos,” then I saw a pen appear and disappear from the reporter’s hand.
Mistakes keep happening. The new feature is dropping fast. Friends were delighted and shocked at the authenticity of the swag for “hip-hop artists cozy Christmas sweater” – until we saw that the rapper’s gold chain had turned into a golden pony tail on the back, and the deer sweater had eight legs.
Sora’s response to the “funeral procession with circus rescuers” hit the nail on the head… except that the colorful, red-nosed person in the coffin was missing his body.
That’s not to say that Sora won’t have an immediate impact on the motion picture industry. Given the unusual instructions, it can replace much of the standard B-roll often seen in YouTube explainers and corporate training videos. (That’s assuming OpenAI won’t be forced to cease and desist training Sora via internet video without the developers’ permission.)
It is something to say that there is a significant barrier to entry when it comes to creating videos that contain anything out of the ordinary, anything that you try to lie about, anything that Sora isn’t specifically trained for. Weeding out all those mistakes, to the point where we won’t immediately notice, can be a frustrating task.
And perhaps these AI videos that are quickly filled with errors will serve as a kind of mass injection – a small dose of post-truth disease, giving our brains antibodies against AI that can better prepare us for the future epidemic of virtual lies. .
Video AI needs to get on the trend train
I’m definitely not that impressed with the AI ​​after I told Sora to take a new one from Lumieres’ The Arrival of the Train. I asked for a video where there is a train it does actually break through the projection screen at the end, crushing the cinematograph audience.
But Sora couldn’t even reach the first short of 50 seconds, ie way out of copyright and widely available online (including an AI-enhanced version). It attracted a movie called “Arrival of a tal [sic] train,” apparently issued in the year “18965.”
As for breaking the actual fourth wall, forget about it: despite multiple attempts to quickly reset the words, Sora simply couldn’t bark out what I was asking. The projection screen remains the same.
Still, this version of Sora could still be a harbinger of other horror visuals to come – perhaps when AI-powered video technology falls into the hands of a future DW Griffith.
Two decades passed in between The Arrival of the Train and Griffith’s infamous filmThe Birth of a Nation (1915)– the first real blockbuster, a milestone in the history of cinema, which also happened to be a twisted story in recent American history full of racist lies.
Griffith’s film, which was protested at the time by the NAACP, was instrumental in furthering the division and revival of the Ku Klux Klan.
So yeah, maybe Sora’s release is slowly pushing us towards a different post-truth world. But even in a future dominated by AI, bad actors will have to work overtime if they want to do more harm to society than the most dangerous cinematograph information.
Articles
Artificial Intelligence OpenAI