The copyright application for the award-winning AI painting was rejected! The author prompted 624 times to create

Do you still remember the first award-winning AI painting "Space Opera"? Recently it has been in the spotlight again——

The author, Jason Allen, applied for copyright for it, but was rejected by the U.S. Copyright Office.

The reason is that the AI content of this painting is too high.

Allen explained that the painting was not entirely generated by AI. He generated the initial image after at least 624 prompts on Midjourney, then used PS to modify the defects and add new content, and finally used Gigapixel AI to add it. Image resolution and size.

That is to say, the finished product also contains a lot of human labor.

In this case, the Copyright Office made a request to "exclude the AI-generated parts of Midjourney and Gigapixel from the copyright statement."

Allen rejected the request and filed for review.

He believes that the Copyright Office ignored the human creativity required to "enter prompts" when creating works using Midjourney. The creative process is equivalent to the creativity expressed by other artists and should be protected by copyright.

However, the reexamination was rejected again. Allen said that this result was expected, but he was "confident that we will win in the end":

If this situation is established, it will bring more unpredictable problems to the Copyright Office.

There are many similar situations. It seems that the copyright issue of AI creation will continue to be played in the legal ambiguity for a long time...

Copyright can be reversed

Last year artist Kris Kashtanova entered this command into Midjourney:

Zendaya walks out of the gate of Central Park, a sci-fi scene of empty New York in the future...

After hundreds of inputs, an 18-page short comic book emerged - "Zaria at Dawn".

Kashtanova acquired the copyright to the work last September and announced on social media that it meant the artist had the right to legal protection for his AI art project.

But not long after, in February this year, the U.S. Copyright Office suddenly reversed course and revoked the copyright protection for this work.

The U.S. Copyright Office said the comic book was "not the product of a human author" and allowed Kashtanova to retain copyright in the storyline.

After that, Kashtanova turned to Stable Diffusion, which can be redrawn based on existing images. She thought it would be very strange to start with an original work of art and then not be protected by copyright.

Kashtanova plans to submit a copyright application for a new comic she created.

** **###### △ Left Kris Kashtanova sketched, right Stable Diffusion generated

The picture above is the work "Rose Mystery" generated by Kris Kashtanova's self-drawn sketch and input into Stable Diffusion. It was published in "Brooklyn Railway" magazine in May this year.

In addition, computer researcher Stephen Thaler previously wanted to apply for copyright for the AI system he developed.

The Copyright Office rejected an image copyright application submitted by Thaler on behalf of the DABUS system on the grounds that "only works of human authors can be copyrighted."

Thaler also tried to apply for patents generated by DABUS in other countries such as the United Kingdom and Australia, but with little success.

From the above examples, it is easy to see that one controversial point in whether AI-generated content should be protected by copyright is how to define the thoughts and creativity expressed by people in the AI creation process.

The U.S. Copyright Office only looks at the results, and those generated by AI cannot enter the protection circle.

On the contrary, what the creators argue is the process of repeated design of prompt words by humans, modification after AI generation, human guidance of AI creation, etc. The final work is not randomly generated by AI.

As Kashtanova wrote in her application for copyright registration of "The Mystery of the Rose":

Although the work was created using AI tools, it visually expresses Kashtanova’s original thinking. Kashtanova controlled the AI tools, deciding the theme of the work and the way it was presented.

I can only say that this issue is really complicated...

AI music is rushing to Grammy

In addition to images, AI-generated audio is also eager to try, the kind that imitates famous singers.

There is a song called "Heart on My Sleeve", which was sung using AI to imitate the voices and styles of Drake and The Weeknd.

** **###### △Drake on the left, The Weeknd on the right

The producer's online name is Ghostwriter977. The lyrics and music of the song are written by humans, and the "singing" is completed by AI.

According to iety, Ghostwriter977 has submitted the song to the Grammys, which will compete for two awards: Best Rap Song and Song of the Year.

Although the song has been removed from major streaming platforms, it still has a chance to become a Grammy Award candidate.

Harvey Mason Jr., CEO of the Recording Academy, which is responsible for the Grammy Awards, said:

AI-assisted music creation can be submitted, but only those who "make a significant contribution" will actually win.

In addition, Ghostwriter977 also used AI to imitate other musicians. For example, he recently posted a song on Twitter using the voices of 21 Savage and Travis Scott.

What’s even more outrageous is that some AI-generated songs are faked as “leaked songs to be released” and used to defraud money online. Well-known R&B musician Frank Ocean is one of the objects of forgery.

An anonymous scammer is selling AI-forged Frank Ocean songs on Discord for more than $4,000 each. He pretends to be a leak of a song to be released, but in order to increase credibility, he actually released a song for sale.

AI-generated music has had a big impact on genuine music.

Universal Music Group, one of the world's largest music companies, has asked streaming platforms like Spotify to prevent AI from accessing music data. Spotify has recently purged thousands of AI-generated songs from the platform.

According to gizmodo, the key to the problem isn't just the songs themselves. An AI startup called Boomy is using "robot listeners" to spread these songs.

Reference links: [1] [2]