ROMA — OpenAI, the Californian company that amazed the world with ChatGptleft everyone speechless again with Sora, an artificial intelligence capable of generating videos realistic starting from a text. Just as ChatGpt has been trained on billions of writings, the new AI was trained with a huge archive of footage. Based on what he has learned, Sora can transform users’ instructions – so-called “prompts” – into moving images. The result is magical. In one of the videos released by OpenAI, a girl walks on a street in Tokyo. All commands given to the artificial intelligence were respected. “The woman is wearing sunglasses, a jacket, red dress and black boots.” Done. “Other people are moving behind her.” Done. “The asphalt is wet and reflects the neon lights.” Done this too.
THE INTERVIEW Cheli: “The greats of the web are as powerful as states. Good journalism must be protected” by Aldo Fontanarosa 17 February 2024
Also the AI improvised direction is impressive. We start from a total that shows the urban context. Then we cut to an extreme close-up of the woman. And the rest of the metropolis is reflected in the lenses of her glasses. In one minute – the maximum that can be achieved with Sora – only two things betray the artificial nature of the images: a small OpenAI symbol – the “proof” that the content comes from an AI – and the sometimes uncertain steps of the “woman in red ”.
OpenAI, a company with an estimated value of 100 billion dollars, also focused on Italy to launch his creation. Sora – which means “sky” in Japanese – can create a city by the sea, inspired by Amalfi, masterfully captured by a drone. Or a dog among the houses of Burano. Or even a grandmother “in a rustic Tuscan kitchen”. But we are faced with demonstrations. Sora, in fact, cannot be proven. And we therefore do not know whether, once opened to the public, it will respond in an equally astonishing way to user requests. OpenAI is currently evaluating any risks and dangers of this technology. There will be work to do. If used the wrong way, a tool like Sora can dramatically increase deepfakes, those videos in which algorithms make an individual – often celebrities or politicians – say (or do) things they have never said or done. The presence of filters or censorship, which should prevent the generation of violent content, offensive and obscene, is not enough. The pop star knows something about it Taylor Swift, who was recently stripped by Microsoft’s AI despite the security measures taken by the company.
AI is ready to help Parliament: “But it will not replace humans in writing laws” by Pier Luigi Pisa 14 February 2024
It’s easy to imagine, moreover, the misinformation that Sora could fuel if someone found a way to use it clonare Joe Biden, Donald Trump or any other person who will run for the White House or a seat in the European Parliament in 2024.
Sora also worries Hollywood and more generally those who earn their living with real videos. “Please don’t make me homeless” MrBeast, the most followed and famous YouTuber in the world, wrote ironically to Sam Altman, the CEO of OpenAI.
Artificial intelligence, here are the first European rights to protect authors and publishers Alessandro Longo 14 February 2024
But there is little to laugh about. A recent study by Carnegie Mellon University, Pittsburgh, highlighted that to produce an image, AI uses the energy needed to charge a smartphone. How much will you need for a video of a few seconds? Then there is the issue of copyright. What images did Sora “study”? OpenAI claims to have drawn on copyright-free public videos. But the company has faced several lawsuits in the past – one from the New York Times – for using sensitive data.