- October 28, 2022
- AI Projects
While she had begun to remove from her base the images generated by artificial intelligence, the Shutterstock platform has finally announced that it will now collaborate with the OpenAI company to offer its customers artificially created images – in parallel with images provided by real photographers or graphic designers. If Shutterstock has planned a form of compensation for artists, the news has raised some skepticism among professionals in the sector.
Shutterstock is one of the world’s leading image platforms: thanks to a community of more than two million contributors, hundreds of thousands of images are added to the database every week; it currently has over 425 million images and over 27 million video clips. The company just struck a deal with OpenAI to offer its customers the ability to instantly generate images based on the criteria they enter. How? Thanks to DALL-E, the text-to-image program developed by OpenAI — and whose second version was recently released.
But like any artificial intelligence, DALL-E must be trained to achieve the best performance. Shutterstock has thus planned to compensate the authors of the images that will be used for this learning. ”The means of expressing creativity are constantly evolving and developing. We recognize that it is our great responsibility to embrace this evolution and ensure that the generative technology that drives innovation is grounded in ethical practices.” said Paul Hennessy, Managing Director of Shutterstock, in a press release.
Compensation conditions that remain unclear
Note that this announcement comes just a few days after Microsoft’s annual conference dedicated to innovation (Ignite 2022), during which the Redmond firm presented several solutions and tools dedicated to graphic design. It turns out that DALL-E 2 is going to be integrated into Microsoft applications available to the general public like Microsoft Designer (available in Microsoft 365) or Image Creator (in Microsoft Bing).
The second version of DALL-E, unveiled in April, generates breathtakingly realistic images. In addition to an even more precise resolution, the program is now able to make modifications to an existing image from indications in natural language, while respecting the play of shadows and light. It relies on a process called “diffusion”, which starts from a pattern of random dots and gradually changes that pattern into an image as it recognizes specific aspects of that image.
In the coming months, Shutterstock customers will receive direct access to these image-generating capabilities, while platform contributors will be compensated every six months via a compensation fund “for the role their content has played: “Shutterstock also created the framework to provide additional compensation to artists whose works helped develop the AI models. The company also aims to compensate its contributors in the form of royalties when their intellectual property is used. “, can we read in the press release?
The company, however, did not specify what percentage of revenue would go to contributors, or how the contributions would be distributed. So some remain skeptical. In practice, it will indeed be difficult to determine which input data was referenced to create a given output element…
Getty Images advocates editing rather than creation
Faced with this legal vagueness concerning the copyright of an artificially created image, companies have so far been reluctant to use them – and this is also the reason why Shutterstock had sought to eliminate them from its database. The situation may become clearer in the coming months.
A Shutterstock spokesperson said the company would continue to ban the uploading of AI-generated works, and that its collaboration with OpenAI was an attempt to embrace new technologies in an ethical manner. It was concluded that the images used to train the AI come exclusively from the archives of Shutterstock and the company said it wanted to ensure that all contributors are protected and compensated.
Note that one of Shutterstock’s competitors, Getty Images, took the exact same turn: while it banned the sale of images created by an AI last month, the company has just announced its collaboration with the Israeli start-up BRIA. Again, this partnership was motivated by the prospect of being able to offer users the possibility of creating bespoke images, meeting specific needs, and using intuitive tools. But according to company CEO Craig Peters, Getty Images’ approach is different: it doesn’t offer to create an image from scratch but to transform an existing image – which, according to the CEO is a more “ethical” use of AI.
“Shutterstock and Getty are no longer just hosting providers, where creators can sell the rights to their digital artworks made using external text-to-image tools. They can now participate in content creation and marketing themselves by providing the software “, summarizes an article from Register.
Today, photographers and artists fear that their income, already drained by content platforms, will decrease further – customers probably prefer to create on the fly the image corresponding exactly to their needs rather than to carry out a search in the catalog. Adrian Alexander Medina, editor of the literary website and magazine Aphotic Realmevokes a risk of ostracization of professional photographers and illustrators.
Moreover, these artistic creation technologies are fed with such an amount of data that they often end up imitating the style of certain artists. The nuance that distinguishes them from pure and simple plagiarism is still unclear… It is certain that future legal challenges will be faced as these technologies are deployed.
