UNIT.City — місце, де люди працюють... КРАЩЕ! Обирай свій простір просто зараз 👉
Олександр КузьменкоAI Eng
27 March 2025, 14:31
2025-03-27
Just a day after the release of the GPT-4o image generator, the internet was flooded with images that echo the styles of “South Park” and “SpongeBob SquarePants.” Is OpenAI violating copyright?
OpenAI has trained its GPT-4o model to generate images and has made it available to paid subscribers. The model has been shown to be adept at imitating the distinctive style of famous cartoons and anime, which has led social media users to take a liking to the way it converts famous images (including the White House brawl) into different settings. Here’s what intellectual property experts have to say about it.
OpenAI has trained its GPT-4o model to generate images and has made it available to paid subscribers. The model has been shown to be adept at imitating the distinctive style of famous cartoons and anime, which has led social media users to take a liking to the way it converts famous images (including the White House brawl) into different settings. Here’s what intellectual property experts have to say about it.
In just one day, social media feeds have been filled with memes in the style of Ghibli, the iconic Japanese animation studio that created such famous anime as «My Neighbor Totoro» and «Ghost in the Shell».
OpenAI CEO Sam Altman even appears to have generated a Ghibli-style avatar for himself on X (Twitter), presumably using his own GPT-4o image generator.
OpenAI’s latest update comes after Google released a similar AI feature for images in its Gemini Flash model, which also caused a viral moment in early March when people used it to remove watermarks from images.
Users seem to be uploading existing images and pictures to ChatGPT and asking the chatbot to recreate them in new styles. Interestingly, a notable portion of them is dedicated to the altercation between Ukrainian President Volodymyr Zelenskyy and US President Donald Trump in the White House. Users recreate photos from it in the style of LEGO, the animated series Rick and Morty, South Park, and many others.
New tools from OpenAI and Google make it easier than ever to recreate the styles of copyrighted works by simply typing in a text prompt. Together, these new capabilities for AI imagery are rekindling concerns that have led to several lawsuits against developers of generative AI models. If these companies train on copyrighted works, are they infringing on copyright?
Evan Brown, an intellectual property lawyer at the law firm Neal & McDevitt, told TechCrunch that products like the GPT-4o image generator currently operate in a legal gray area. Brown said the style is not protected by copyright, meaning OpenAI isn’t breaking the law by simply generating images that look like Studio Ghibli films.
However, as the lawyer notes, it is likely that OpenAI achieved this similarity by training its model on millions of frames from Ghibli anime and other media. Even if that is the case, numerous courts are still deciding whether training AI models on copyrighted works falls under fair use protection.
«I think it raises the same question that we’ve been asking ourselves for a number of years. What constitutes copyright infringement? What are the consequences of copyright infringement when we go online and copy into these databases?», — Brown said.
The New York Times and several publishers are pursuing active lawsuits against OpenAI, alleging that the company trained its AI models on copyrighted works without proper attribution or payment. Similar claims have been made in lawsuits against other leading AI companies, including Meta and AI imaging startup Midjourney.
An OpenAI spokesperson said that while ChatGPT refuses to copy «the style of individual living artists,» OpenAI allows it to copy «broader studio styles.» It’s worth noting that there are living artists who are considered pioneers of their studios’ unique styles, such as Studio Ghibli co-founder Hayao Miyazaki.
TechCrunch tested several popular AI image generators, including Google’s Gemini, xAI’s Grok, and Playground.ai, to see how well they matched Ghibli’s style. The publication says that OpenAI’s new image generator created the most accurate copy of the animation studio’s style.
In a post on Wednesday, CEO Sam Altman acknowledged that the image generation tool had proven more popular than he expected, adding that «the rollout of our free tier will unfortunately be delayed for some time».
During the presentation, Altman said that native GPT-4o image generation is already available in ChatGPT and Sora for Pro subscribers for $200 per month.
Як нейромережі бачать вільну та незалежну Україну? Тест dev.ua
Нейронні мережі для генерації зображень бачать світ по-своєму, їхню логіку зрозуміти часом зовсім неможливо. Але таки хочеться. На честь Дня Незалежності України редакція dev.ua вирішила провести невеликий експеримент.
Ми задали чотирьом різним нейронним мережам п’ять однакових запитів: «прапор України», «День Незалежності України», «український Крим», «перемога України» та «українці». Отриманими результатами ми ділимося з вами нижче.
У TikTok тепер можна генерувати фон за допомогою нейромережі. Ми протестували її та ділимося результатами
У TikTok з’явилася нова функція «Розумний фон». З її допомогою як фон для тіктоків можна підставляти згенеровані нейромережею зображення. Редакція dev.ua протестувала цю технологію і ділиться своїми враженнями.
