Getty Images has announced its entering the lifeless world of AI image generating with a ‘commercially safe’ tool unimaginatively named ‘Generative AI‘.
Computer parts manufacturer, NVIDIA, announced a partnership with Getty back in March 2023 to develop two generative AI models. This appears to be the first.
Getty explains Generative AI is ‘trained on the state-of-the-art Edify model architecture, which is part of Picasso, a foundry for generative AI models for visual design’. The press release fails to elaborate what Edify and Picasso are, suggesting it’s either not interesting or average folk aren’t smart enough to understand. Or they expect journalists to do extra leg work!
NVIDIA’s Picasso is apparently available for software developers to train AI models on their proprietary data. Adobe, for instance, utilised Picasso to train AI models ‘to accelerate workflows’, and Shutterstock is developing 3D models trained on its content.
Getty is differentiating itself by marketing the Generative AI tool as ‘commercially safe’. It has been trained on the Getty library, ‘with full indemnification for commercial use’ and no risk for nasty copyright litigation.
‘We’re excited to launch a tool that harnesses the power of generative AI to address our customers’ commercial needs while respecting the intellectual property of creators,’ said Craig Peters, CEO at Getty Images. ‘We’ve worked hard to develop a responsible tool that gives customers confidence in visuals produced by generative AI for commercial purposes.’
Getty’s text-prompt visuals come with a standard royalty-free licence, which includes the right to perpetual, worldwide, nonexclusive use in all media. The content will not be added into Getty or iStock’s content library for others to licence, and Getty contributors will be compensated when their content trains the AI. It’s unclear how much, or how Getty determined the contributor payment rate.
‘We’ve listened to customers about the swift growth of generative AI – and have heard both excitement and hesitation – and tried to be intentional around how we developed our own tool,’ said Grant Farhall, chief product officer at Getty Images. ‘We’ve created a service that allows brands and marketers to safely embrace AI and stretch their creative possibilities, while compensating creators for inclusion of their visuals in the underlying training sets.’
While AI image generators threaten the future of stock imagery and graphic design, they also need to be trained on large image datasets.
Getty, with one of the largest and most varied image libraries, is in good step to train an AI image generator without committing potential copyright infringement by scraping data from millions of online pictures. And Getty may also sell access for other developers to train their AI on its dataset.
But rogue AI developers may train their AI models by quietly scraping image data without permission from wherever it’s available. It’s yet to be tested whether it even constitutes copyright infringement, although Getty aims to soon find out. Earlier this year Getty filed a copyright lawsuit against Stable Diffusion, alleging the open-source text-to-image model unlawfully utilised up to 12 million images from the Getty archive. Part of Getty’s evidence is how the Stable Diffusion AI model was able to re-produce the Getty Images watermark.
Getty has also co-signed an open letter, along with several major media companies, urging lawmakers around the world to introduce regulation that requires transparency into training AI datasets. The letter also calls for consent from rights holders to train datasets, as well as requiring AI users to ‘clearly, specifically, and consistently identify their outputs and interactions’.
Stock agencies may emerge as key players in the battle to commercialise and fine tune the AI image generator. Adobe’s Firefly tool is part of an Adobe Stock subscription, and Shutterstock has incorporated OpenAI’s Dall E 2 into the platform.
Be First to Comment