The AI image war is actually a copyright dispute

Adobe joins the fray

Designers who boycotted AI painting a few months ago are clearly unable to stop the torrent of technology. On Tuesday night, local time, content productivity tool manufacturer Adobe announced the launch of a creative generative AI called "Firefly", officially entering the AIGC commercialization track.

As a developer of content production software such as the retouching tool Photoshop and the video editing software Premiere, Adobe's choice of AI track entry point is also related to its old line - image generation. In addition to generating images from text descriptions, you can also generate artistic fonts, recolor target objects, and more.

In addition, Firefly generates images based on peripheral content, which allows users to test and refine images more effectively. In graphic design and illustration work, custom vectors, brushes, and textures can be generated from a few words or even sketches, with the ability to edit content created with familiar tools.

Just as Microsoft stuffed GPT-4 into the office bucket to improve the efficiency of office software, Adobe also plans to cram the AI tool Firefly into the Adobe family bucket to make it easier for designers to use. At present, the series of tools has been integrated in Adobe's enterprise-level creative tool Adobe Express, and will be fully integrated into its Photoshop, Illustrator and Premiere.

For Adobe, who sold software, this AI also has a killer feature - commercial copyright. Most AIGC tools on the market today have a common problem, and do not claim where the image library they use to train AI comes from, which may contain a large number of unauthorized images. For business users, using such AI to generate pictures may have unknown legal implications.

Ely Greenfield, Adobe's chief technology officer, said that when training Firefly, Adobe carefully selected data sources, mainly including Adobe's own copyrighted images, publicly licensed pictures and expired copyright images, so the trained AI is commercially safe.

Help, not replace

AIGC has always been a highly controversial topic in the circle of art creators. Some see it as a new tool that can be exploited, others as stealing and putting one's job in jeopardy.

There are also image copyright holders who believe that the AIGC model uses the pictures in its own database without authorization, which is an infringement. For example, Getty Images, one of the world's largest image library services, is now fighting an AIGC unicorn and Stability AI, the parent company of Stable Diffusion, accusing its engine of improper training using copyrighted photos.

A person working at a picture copyright company told Beijing Business Daily that despite the efforts of large commercial companies to promote the implementation of AIGC, its legal and copyright ownership disputes still need to be resolved. On the one hand, the training data used by the "big model" used by AIGC may come from copyrighted works and is suspected of infringement. In addition, the commercial use and copyright attribution of AIGC works still need to be reached by the society and further regulated by law.

According to Chen Jia, an independent international strategy researcher, the industry has also creatively proposed a series of solutions to the possible plagiarism problem of AIGC works. For example, the anti-tracking GPT application tool developed based on GPT4.0 is used to identify whether GPT is used extensively in art exam works. Firefly is also a development product under this framework, and Adobe has always had a global comparative advantage in the field of property protection. In general, as long as the supervision is properly regulated and properly applied, technological progress will not only not harm intellectual property rights, but can be protected by more advanced technological means.

In addition to copyright issues, many creators have also raised doubts. As a company focusing on the field of creative design, Adobe provides many shortcut tools in the continuous upgrading and evolution, such as Photoshop provides at least 4 kinds of cutout tools, under the intelligent standard, they make the work of "cutout" very easy and simple. Designers believe that even Adobe has begun to use AI to generate images, and the design field still needs real designers?

In fact, before that, there were many AI painting tools appearing. In August 2022, at the Colorado Art Fair in the United States, a painting called Space Opera won first place in the Digital Art category. But it soon became clear that the drawing was not a real person, but had been generated by the game designer using an AI tool called Midjourney. Previously, NetEase's creative community lofter also caused controversy, because the official provided the function of AI painting, which made creators in the community feel offended.

Art designer Feifei told Beijing Business Daily that software such as FireFly will indeed make it easier for creators to design commercially, but in a sense, it is also erasing the creative thinking of creators, which may be aggravated after Firefly enters Photoshop, Illustrator and other software.

In this regard, Adobe believes that AI is providing a new way to open up the world, and designers can realize ideas in the most convenient way, "helping creative people, not replacing them."

The night the comet came

Starting from ChatGPT stirring a pool of spring water, AI-related technologies such as multi-modal and large models continue to emerge, and AIGC has become a new battlefield for technology giants to chase the deer.

"The iPhone moment of artificial intelligence is on." On the same day that Firefly was released, at NVIDIA's annual GTC developer conference, CEO Jensen Huang mentioned this metaphor four times in his speech. Huang believes that under the wave of technology, everyone is a programmer, and generative AI is the new computing platform. Just like PCs, the internet, mobile devices, and the cloud once were, generative AI will reshape nearly every industry.

At the GTC conference, NVIDIA launched four new chips in one go, all of which are aimed at the most popular generative AI. In addition, NVIDIA launched its own service called "Picasso", which uses artificial intelligence technology to generate images, videos and 3D applications from text descriptions.

While Huang was showing his strength, several major technology companies were also moving forward. Among them, Google opened its internal beta application for Bard, a product benchmarked against ChatGPT.

This isn't Bard's first appearance. As early as February 8 this year, Google has shown this chatbot. But that presentation caused Google's stock price to plummet 9% that day, losing $100 billion in market value in one day, and Bard's test for the public was forced to be postponed.

A month later, Bard launched the closed beta application again. However, it doesn't support more languages, including Chinese, or responses to code yet. Some testers who got the internal test code said that "the dialogue effect is still far worse than ChatGPT".

Microsoft, on the other hand, has updated the preview of the new Bing and Edge browsers with three new features: Bing Image Creator, AI-powered Stories, and Knowledge Cards 2.0. Among them, the most striking is the Bing Image Creator, which can draw, making the Edge browser the first browser to integrate an artificial intelligence image generator.

Today, technology iteration has become so fast that humans are terrified. As Microsoft founder Bill Gates wrote on his blog GatesNotes, the era of AI has begun, and the AI revolution is no less important than the birth of mobile phones and the Internet, which is the only two revolutionary technological advances in his lifetime.

Source: Beijing Business Daily | Written by Fang Binnan, Zhao Tianshu

Editor: Gao Shanshan

Process Editor: Guo Dan