Last May, actress Scarlett Johansson published a statement on the publication of a new voice for the Chatgpt 4.0 bot by Openai called “Sky”. For many – even for their “closest friends”, as Johansson said – the voice was not to be distinguished from their own.
That would have been worrying enough, but it came even worse: the technology giant had asked her to set “Sky” himself months earlier, hoping to partially use their celebrity in order to be able to partly use the tensions between Hollywood and the industry for artificial intelligence to reduce. She said, she said in her statement.
Johansson said that she was “shocked, annoyed and amazed” when she heard the demo that was published, while the CEO of Openai, Sam Altman, asked her to consider an official partnership. Altman even tried to consciously create the Johansson effect by posting the word “her” on X (formerly Twitter) before the start, which seemed to be an allusion to the science fiction love film “Her” from 2013, in which Johansson speaks an intelligent chat system.
With a legal advisor at her side, according to Johansson, she pushed the company to disclose the specific steps to program “Sky”. Thereupon had “reluctantly” agreed to remove the function.
Raub train of the AI companies
Welcome to a new normalcy for prominent personalities of public life: You not only have to worry that random fraudsters with the help of AI generated voices or pictures than they spend (Brad Pitt was dismayed by the history of a French woman who was more than 850,000 dollars lost to a fraudster who set up an Instagram account with AI generated representations of the actor in the hospital), but also that a billion-dollar company from Silicon Valley could try to “borrow” their valuable portrait to advance their corporate goals .
Since your bots are already being trained with private data and copyrighted material, why should you stop when using famous faces?
In this context, a prominent could be relieved to hear from Loti, a company for “similarity protection technology”, which last autumn has raised a 7 million dollar starting capital round and continues to grow. And although co -founder and CEO Luke Arrigoni cannot call specific customers who are represented by heavyweights like William Morris Endavor, Creative Artists Agency and United Talent Agency, he can say with a clear conscience that he was consulted with many top -class incidents where an actor Or musicians suspected that its unique properties in one form or the other were illegally copied.
If it appears that non -licensed images, videos or audio files were created based on the model of a living person, Loti will probably contact this person by the management team. “We are almost always switched on in situations in which you hear that a prominent is either fake or especially used in a large AI product,” says Arrigoni Rolling Stone.
Loti’s work can include a technical analysis of the content to help lawyers understand the likelihood of whether a deep pap is based on your client. “For example, one could say to someone that this voice matches this or that percentage,” explains Arrigoni. “If you sing a 10- or 15-second A cappella segment, I can find out when people generate this voice with the help of AI in a different song.”
In addition to such evidence of a possible misuse, he can prepare the lawyers’ teams for the arguments that they will probably hear from the lawyers of the AI companies who defend their products and content. However, he quickly points out that it is his job to provide you with relevant information, “not to tell you what to do”.
“Euphoria” as an impetus
Arrigoni actually has little in the hat with court halls, but is part of data science (he is a self-proclaimed “math type” and loathes the more trendy keywords of the AI scene). Before that, he worked on data questions for large brands for a decade. Loti was first created in 2022 as a kind of non -profit organization when he and his wife, co -founder Rebekah Arrigoni, saw the second season of HBO’s “Euphoria”, which dealt with revenge pornography.
At that time he worked with a customer who was able to carry out facial recognition comparisons with only a part of a person’s face. The couple had the idea of using this technology to report cases in which the image of a person appears online without their permission, and then send copyright information in order to have them removed.
“We thought: ‘Why don’t we use this technology and then focus on the Internet? We can find out where people don’t belong and we create a script to cause the distance, ”recalls Arrigoni. “That was something that we both were passionate about.”
When Arrigonis Tech team developed this AI tool to search the Internet and Rebekah began to develop an advisory side, the Hollywood strikes 2023 began, with AI being a focus in the industrial action between actors and authors and the management of the studios.
“I called friends from the entertainment industry and asked: ‘Hey, would you like to use the thing that I developed to find and ward off AI imitations in this unique, specific situation?'” Says Arrigoni. They were actually interested and Loti “massively expanded the scope of what we scanned,” he says. “It was not just about a few problem areas on the Internet, but about the entire Internet, and we also added the voice. We had a Deepfake detection. “
What Arrigonis now has a pioneering role in a wild new border area of digital security and data protection and advertises that he can “search for 100 million pictures and videos per day algorithmically to the abuse of their content or image”, content with the face and the face and Voice of a customer compares and immediately sent notifications about the removal of copyrights.
LOTI is boasting that its automated extinguishing machine has a success rate of 95 percent within 17 hours and most of the violations within a calendar day – which means that this material is usually disappeared before it is made accessible to the general public In contrast to the Johansson-like Openai voice “Sky”.
Arrigoni says that you have effectively automated a kind of work that was done until recently by a far less efficient and incomplete process: Fans noticed an unauthorized use of the image of her favorite star and sent it to the management of this person, whereupon lawyers And agents negotiated with the platforms to remove the hurtful content.
Celebrities have to proceed very quickly against Deepfkes
That is simply not enough in today’s accelerated surroundings, says Arrigoni. “You can hire your law firm to carry out 15 takedowns today, and 14 new things are uploaded,” he explains. Loti, on the other hand, is able to keep up with the tide. “Today I can arrange a thousand takedowns for all 985 things that were uploaded yesterday, and a few things are uploaded in the morning,” he says.
Ideally, the fans don’t even have the chance to encounter non -licensed content. In the meantime, famous customers can request specific filters to intercept the material that harms its reputation most, such as: B. Explicit Deepfake videos.
Of course, it is a bit tricky to convince entertainer to convince that AI is the solution to the AI problems that the creative landscape has hit in recent years. Arrigoni has to convince her that Loti fought fire with fire. “It is so embarrassing when you say, ‘I have a AI company’,” he jokes, adding that he often ‘apologizes’ for the technology. “Many of these AI companies tend to make themselves bigger than they are-making the world a better place, things like that. We don’t try to pretend we are like that. “
Arrigoni says potential customers: “Technically speaking, we are trying to prevent anyone from using AI to use their income, their creativity, their brand, their reputation.” Ultimately, he says are Many impresses how much LOTI can catch on the net, although they don’t just want to protect themselves. “One of the greatest incentives for our product is that you don’t necessarily benefit from it if you pay us for it, but you are tired of being cheated.” (Again our condolences to the woman who thought she would be for Pay brad pitts kidney treatment.)
Loti is not only used by Hollywood
In addition to well -known actors, LOTI also works with high -ranking military, politicians, CEOs, estate administrators and a number of musicians from Nashville, which means that new challenges and precedents will be created in a variety of areas.
What happens when Dreamworks works with a Ki startup and creates an animated film figure that looks confusingly similar to a celebrity that is not involved in the project? Will some of the citizens of Florida, the supposedly template for scenes in the upcoming “Grand Theft Auto VI”, will unite to sue Rockstar Games? And the states act quickly enough to enact laws such as the Ensuring Webess Voice and Image Security (Elvis) Act by Tennessee, which was signed in a groundbreaking attempt last year to prevent the voices from being cloned and fake illegally ?
Whatever may come, Arrigoni wants Loti to continue to ensure that pop-up “fakes” are no longer “economically viable” as an Internet fraud: “Make it more complicated for the developers because it is too easy for them at the moment – That is the part we want to change, ”he says.
Since the arms race continues in the field of artificial intelligence and other competitors will get into the game with full strength, he can hardly imagine that it will be boring in the future. But at least one question has already been clarified: Your chat bot assistant will not sound like Scarlett Johansson.