metamorworks - stock.adobe.com
Stock image vendor Getty Images' new generative AI tool for small and large businesses is the latest entry into the market for image-generating systems.
The Getty platform, Generative AI by iStock (a division of Getty), is powered by Nvidia Picasso, a foundry for generative AI models for visual design. The image generator is trained exclusively on Getty Images' library and engineered to guard against copyrighted elements, according to Getty.
"We know everything that was in the training set. And therefore, we are confident that it can't produce visuals that would violate trademark or copyright or create any other legal risk for customers," said Grant Farhall, chief product officer at Getty Images.
To provide confidence to customers, Getty is offering users standard $10,000 legal indemnification coverage for any image and video generated with Generative AI by iStock.
The image-generating tool, introduced on Jan. 8 at the CES consumer electronics conference in Las Vegas, comes after Getty released its own AI-powered image generator in September.
Compared to the AI-powered image generator, Generative AI by iStock is meant for small to mid-sized business users. The Getty-branded AI-powered image generator is aimed at larger enterprise use.
Learning from copyright lawsuits
The new Getty offering also comes as numerous copyright lawsuits have emerged involving copyrighted material used to train generative AI systems.
The New York Times recently filed a lawsuit against ChatGPT creator OpenAI for allegedly using copyrighted material to train its systems.
Also, Getty launched a lawsuit against Stable Diffusion creator Stability AI. Getty accuses the AI vendor of using more than 12 million of its photos to train its system.
Generative AI by iStock reflects lessons Getty is learning from its suit against Stability AI, said Liz Miller, an analyst at Constellation Research.
"[It has] built in a lot of safeguards around training data to ensure that anything generated by its model is safe for use," Miller said.
These safeguards are like some of the protections Adobe has built into Firefly, its generative AI imaging system.
Adobe and Getty offering these kinds of protection is telling of the problems many enterprises face when considering generative models and tools, Gartner analyst Issa Kerremans said. Part of that predicament is choosing an imaging model that is more photorealistic and higher resolution, such as Midjourney V6, or a safer but perhaps adequate model from one of the established vendors, she said.
"Our clients face a dilemma," Kerremans said. "Opt for an ethically trained model like Adobe's and now Getty's … or choose a more controversial option like Midjourney V6 for incredibly realistic outputs at the cost of legal peace of mind."
Issa KerremansAnalyst, Gartner
However, the option of legal protection offered by vendors such as Getty and Adobe provides enterprise users with an added layer of protection, Kerremans said. Such legal indemnification for users of generative AI products is becoming widespread, but it is unclear how effective it is.
"This brings individuals to weigh the legal foundation of their AI choices against the allure of cutting-edge realism," she added.
As the controversy surrounding AI image generators grows, tools such as Adobe and Getty could become the preferred option. It is also possible that copyright regulation could move in the direction of the approach in Japan, where copyright is not enforced for the training of generative AI programs, Kerremans continued.
"The landscape is dynamic, and the future of AI image generation will likely hinge on finding the delicate balance between realism and ethical considerations in a legally compliant manner," she said.
Competing against Adobe
While Generative AI by iStock could provide users peace of mind, its ability to compete against the likes of Adobe Firefly is yet to be seen, said William McKeon-White, an analyst at Forrester Research.
"Generative AI is kind of hard," McKeon-White said. With the new developments in generative images and video and its use in marketing, it would be interesting to see how this model performs in those areas, he added.
New developments include the use of generative AI tools in image design for marketing. An example is using generative AI tools to manipulate a photo's depth of field instead of doing it manually like in Adobe Firefly.
"It will be interesting to see how effectively [Generative AI by iStock] is able to make itself useful to individuals and to small businesses," McKeon-White said.
While Generative AI by iStock could be a better option than OpenAI partner Shutterstock's image-generating mode because of its copyright protection features, it might not be as good as Adobe Firefly, and neither does it need to be, Miller said.
"It is a case of just slightly different applications for slightly different markets with slightly different budgets and uses," she said. CMOs will pick the best tool and application for the application, she added.
Moreover, the iStock model is meant for a small business owner or user who wants to create a safe, reliable, customized image, Miller said.
Generative AI by iStock costs $14.99 for 100 generations.
Esther Ajao is a TechTarget Editorial news writer covering artificial intelligence software and systems.