Disney and Marvel Studios' TV series "Loki (Season 2)" officially launched on the streaming media platform last Friday. Although the series itself has a good reputation in reviews, a poster on social media platforms triggered an "internet investigation" about generative AI.


(Promotional poster, source: Disney, Marvel, TheVerge)

Since the series itself is related to "time travel", it is reasonable for the clock style to appear on the promotional poster. However, in the eyes of some professional designers, the AI ​​elements in this poster made them feel stuck in their throats.

Graphic designer Katria Raden pointed out on social mediaIn the medium-spiral clock in the background of the poster, obvious AIGC characteristics appear, such as randomly generated meaningless distorted graphics.


(Source: Disney, Marvel, TheVerge)

Subsequently, savvy netizens found pictures that appeared to be the background of the clock on Shutterstock, a well-known copyrighted image platform. Some netizens spent several dollars to buy this picture.Although the metadata of the image does not make it clear how the image was generated, the same contributor uploaded a large number of similar spiral images, raising the suspicion that it was generated by AI. Since this picture was uploaded this year, it objectively excludes the "alibi" that the technology in earlier years was insufficient to support the generation of such pictures.



(Source: Shutterstock)

At the same time, there are also multiple AI detection tools showing that this picture was generated by AI. It should be noted that since there are currently no AI detection tools that can achieve high-precision judgment, these conclusions can only be used as a reference. In addition, since the postures of many of the shrunken characters in the poster are also quite stiff, it is not clear whether this poster has a higher "AI content".

Whose "pot"?

Like Adobe, the “data rights” concept stock that investors are very familiar with, Shutterstock also has restrictions on AIGC, such asThe platform stipulates that except for images generated by self-operated AI tools, other AI images must not be posted for sale on the platform to ensure that the platform has the copyright for the images sold, and AIGC images also need to be clearly marked. Shutterstock has also emphasized that because the company's AI tools are trained using copyrighted image libraries, they can be used safely for commercial purposes.


However, in the face of this controversial image that plunged Disney into controversy, Shutterstock did not publicly respond to the query, nor did it clarify how the platform takes action to ensure that sellers selling images comply with AI rules.

Of course, the relationship between Disney and AI has actually gone beyond the level of poster production. In another Marvel work "Secret Invasion" in June this year,Because the use of AI to create title titles has caused shock in Hollywood. At that time, Disney also specifically emphasized that the use of AI technology did not result in the reduction of designer positions in the project.


(Source: Disney, Marvel)

Katria Raden commented that for many hard-working artists, selling licenses on picture websites has always been one of the ways to make a living, and now this AIGC technology is based on "large-scale exploitation and wage theft."

The well-known technology media TheVerge commented,Since the copyright holder himself did not indicate the source, Disney may not be aware of the status of this image. But even so, allowing random patterns to appear in public materials is not a good sign.

This latest controversy also reflects the simmering conflict between content creators and the AIGC industry. In the past year, many image copyright providers have launched AI tools, focusing onIt allows users who lack creative experience to generate large-scale works of art, and it is also cheaper.

More importantly, with the rapid evolution of AI, some obvious AIGC characteristics are being quickly corrected. For example, DALL·E·3 released by OpenAI last month has been able to solve the problem of text being unable to be accurately generated, which will also make it exponentially more difficult to distinguish AIGC works in the future.