Laurent - stock.adobe.com
Executive interview: Balancing AI with human creativity
We speak to the chief product officer at Getty Images and iStock about the role generative AI can play in the image-making process.
Getty Images has worked with graphics processing unit (GPU) and artificial intelligence (AI) specialist Nvidia for several years now, helping the firm understand the quality of its data models against Getty’s library of visual content.
“Nvidia shares our belief in terms of respecting intellectual property and upholding artist rights,” says Grant Farhall, chief product officer at Getty Images and iStock. “We share a perspective on the problems we’re trying to solve for our customers, and we have complementary skills and expertise.”
He says Nvidia has some of the best data scientists in the world. “They’ve also got GPUs, which have become a very precious commodity in this new world,” says Farhall.
Farhall has worked at Getty Images for 13 years across various product and ecommerce roles, and currently leads the overall product strategy, which he says “obviously has a lot of elements of AI now”.
“We look at technology as a way of helping our customers to be connected to the right visual content so they can amplify their creativity, save themselves time and money, and manage their risk,” says Farhall.
With this in mind, he says Getty Images looked at where AI can help. The company recently added “refine” and “extend” features to its Generative AI by iStock platform.
Among the questions he says Getty Images asked when developing its AI platform were: “Where does AI present opportunities to solve problems in different ways? Where does it present the ability to solve problems that were previously unsolvable? How do we deliver the value of AI to our customers? And how do we then develop a product that aligns to certain tenets?”
Read more about GenAI
- The integration of Copilot into the Mural visual collaboration platform shows how vendors are working together to bring customers generative AI tools that interoperate between systems.
- The social media giant's decision to label AI-generated content is a good first step, some observers say, but does not eradicate the problem of disinformation in an election year.
The principles the company wanted to uphold included being supportive of the advancement of AI in a way that respects intellectual property (IP) rights, protects and upholds artist rights and sustains what Farhall describes as “the creative ecosystem, including the photographers, illustrators and videographers that create the amazing work we licence through our library”. But with AI, he says it’s important to recognise these works as training set data for AI models.
Looking at the prospect of AI emulating real photographers, illustrators and videographers, Farhall says: “Our belief is that human-created content shot on camera will stand above what AI can do.”
He believes this is because AI is always going to involve looking back at datasets. “It’s going to be learning and training itself on the basis of a corpus of content that reflects the world as it was, whereas human creative content is going to be more contemporary, more authentic and more of a look forward.”
As an example, Farhall says: “If we rewind a few years when the pandemic happened, instantly what everything looked like changed. What did work look like; education; socialisation? The visualisations of what the world looked like changed across all those different concepts and we activated our contributor base to provide contemporary, timely and accurate visualisations of what that world looked like at that time.”
He says AI would not have been able to do this as it can only draw on the images it has, which, at the time, would have been based on a pre-pandemic world. “AI would still be basing its visualisations off everything it had trained on up until that point,” says Farhall.
AI as an additive
He strongly believes there’s a place both for content shot, drawn or painted by people and AI-generated content, and also sees opportunities for AI content to be additive as part of the human creative process. “I certainly have seen that when you put these tools in the hands of talented creative people, it does allow them to stretch their imaginations and think about concepts differently,” says Farhall.
“If you were writing a story about climate change and trying to figure out a visual element, a really common image is the polar bear isolated on an ice flow surrounded by a lot of melted water,” he offers as an example. “But that’s kind of overused. [With AI] you could create a visual with hundreds or thousands of penguins marching through a city street.
“It’s a very different way of illustrating climate change, but it’s impossible to shoot unless you have access to a colony of penguins and could fly them and convince the city authority to let you march them down a street.”
Farhall says brands need to be thoughtful about when to apply AI visuals into their work, and how they make those decisions based on the message they want to put out and the audience they want to reach.
The implications are that if an AI-generated work is passed off as something created by a human, it may have a detrimental impact on how the brand is perceived in the same way that a heavily manipulated image is considered artificial. But this entirely depends on the context and how the visualisation is intended to be used.
Overall, Farhall sees a huge opportunity in using AI to drive visualisation creativity. “I’ve seen examples where the longer visual content creators have access to AI tools, the more they build muscle memory around them and start to get really imaginative,” he says. “One of the main things we’re interested in is where these tools can provide the ability for someone to safely create something that is out of reach.”
