Understanding GenAI’s Role in the Enterprise Content Creation Industry
As part of AiThority.com’s IWD celebrations, we interviewed Sr. Director AI ML of Getty Images, Andrea Gagliano. Recently, Getty Images partnered with NVIDIA to build endless Gen AI possibilities in content and media. Andrea discussed the Getty-NVIDIA partnership and the changing role of women leaders in the AI industry. Our conversation majorly revolved around the quality of AI training, AI biases, and the use of AI at Getty Images.
Here is the full interview with Andrea.
Hi Andrea, welcome to our AiThority Interview Series on the occasion of IWD. Please tell us about your AI journey.
I have been with Getty Images for more than six years, becoming Senior Director of AI/ML and working to build the company’s AI offerings. In 2020, I founded the Getty Images AI/ML team, which has grown to 20 people at various levels, half of whom are women.
My AI/ML journey began in 2015 while I was a graduate student at UC Berkeley studying machine learning and creativity. While there, I worked on an award-winning computer-generated poetry project that grappled with ideas of human and machine collaboration in the artistic realm, authorship, and AI’s ability to inspire young creators.
Outside of work, I develop art that introduces AI to non-technical people in a way that sparks conversation and discussion about AI and how we want to approach technology as a society. An example of this is Madam President, which I created in 2020 to illustrate what the Presidents of the United States would look like if they were women.
What significant AI generation tools have Getty Images developed over the last few years?
It certainly has been an exciting time here at Getty Images and iStock! In September 2023, we announced Generative AI by Getty Images, which is a tool that leverages the company’s best‑in‑class creative content to create commercially safe AI-generated images. Powered by NVIDIA, Generative AI by Getty Images provides indemnification for commercial use, giving users the confidence to create unique visuals that bolster marketing, social media, and website content.
Quickly following that, we launched Generative AI by iStock in January 2024. This tool is geared toward small businesses that typically do not have the same amount of resources as those in the enterprise space. It provides the same powerful AI generation capabilities as Generative AI by Getty Images and iStock’s standard indemnification, giving small business owners the peace of mind that they are not at risk of copyright or intellectual property infringement.
We’ve also just announced Refine and Extend.
Refine allows users to easily add, replace, or remove selected elements from the images created by the AI generators.
Extend allows users to expand an image in multiple directions and automatically fills in the background.
What sets Getty Images apart from other service providers in the same industry?
Our generative AI tools are trained from our creative library, including exclusive premium content, meaning all materials used in the training data is released content.
Not all AI tools on the market are trained in the same way, and many are not transparent in how they use data sets to train models. Many generative AI tools use the same data sets that are used to train other AI models.
When information is simply pulled from the internet, it creates a situation where the AI-generated output is riddled with copyright issues and/or intellectual property concerns.
Because our tools are trained exclusively on our creative library, they do not know of specific celebrities, corporate logos, or other trademarks. They cannot create visuals with these elements, therefore providing our users with unique, commercially safe images that can be incorporated into their content and campaigns.
Additionally, images generated by our tools are not added to the library to be licensed by someone else.
Other companies in the industry do not put as heavy of a focus on transparency and do not share the same sense of urgency when it comes to implementing regulations around identifying AI-generated images.
For example, there is an “AI Generated” watermark on the preview of images generated by our AI tools. The metadata on AI-generated images also indicates that the images were created using a specific AI model and model version.
As our customers continuously think through how they should best deploy AI tools across their visual needs, these steps are to help ensure that there is no confusion as to where and how the image originated.
Our tools are also different because others on the market limit indemnification to enterprise clients or only at a specific request. This puts a significant risk on small business clients that have limited resources and capabilities to license the content or make sure that the images used do not violate existing protections.
Our VisualGPS research recently found that 42% of small businesses are using AI-generated content to support their marketing efforts, proving that this type of protection is necessary as more small businesses explore generative AI.
Biases in data sets directly correlate with AI-generated content filled with stereotypes and misrepresentation. How do you work to make sure that biases are removed from data so that the content is accurate?
While biases in the data sets are impossible to remove completely, there are steps that teams can take to make sure that they are continuously working to counteract them. This requires careful thought at every level, from coming up with the training set to building the model architecture. It is necessary to continually revisit the models to make sure they are always trained on the content not already protected, and when protected content is spotted in the model, steps are taken to immediately re-train so that it forgets those protected images.
What factors contribute to the difficulty of understanding and deploying AI for operational effectiveness?
Generative AI is incredibly complex.
There are multiple layers to not only how it is built, trained, iterated based on the dynamic industry, incorporated into various business strategies, and implemented into daily practice.
Before Generative AI by Getty Images launched, we had worked with NVIDIA for more than three years to make sure that our research and model training was accurate and trustworthy.
Now, we constantly look at our models to make sure that they evolve in the right way so that our users can rest easy knowing that their AI-generated visuals were created responsibly and in ways that minimize risk for their business.
From the New York Times’ list of the top AI leaders, there is a notable lack of women on boards for leading AI companies as well as throughout the industry. Why do you think that is?
In my experience, there are certainly fewer women working in AI than in other fields I’ve been in, like data science insights, because it is a relatively new field. AI is quickly moving out of academia and research labs, and into businesses and products. Both data science insights and AI have an under-representation of women, but the number of positions available in academia and research labs is limited. Those who want to be involved need to know the right people who can open the right doors.
As more budgets are invested in AI and businesses incorporate it into their overall strategies, more opportunities will emerge, giving more opportunities for women to lead.
There has been a bigger push around women in tech, and there are many opportunities for women to get involved in AI given the ability to move into roles and learn as they go considering how fast the industry is evolving.
Top Gen AI News:
Fujitsu, Carnegie Mellon Develop AI Social Digital Twin with Pittsburgh Traffic Data
I think as more women step into leadership roles and begin to lead the AI charge, other women will follow and begin to chart their own AI paths.
You’ve led a diverse team at Getty Images for a while now, and have noted that collaboration is a major benefit to your success. How would you advise other leaders to prioritize such an approach?
I’ve found incredible benefits in emphasizing the goals and successes of our team instead of individual goals.
The AI industry is enormous and changes at a very rapid pace. It is more than a full-time job to keep up with it all; no one person knows everything or has all of the answers.
Part of that is promoting a focus on curiosity – we empower our team to ask questions and knowledge share instead of focusing on any one person’s success. I also believe it is important to empower our team with the confidence to take risks and accept that it is OK to fail as long as we collectively learn from those instances.
For example, when one of our team members shares the results of a proof of concept they built, I like to make sure that questions asked during that conversation come from a place of curiosity, a place of improving our knowledge.
I like to challenge assumptions and challenge the team to improve the next time around, which helps inspire others on the team.
I truly believe that the sum of the parts is stronger than any one individual, and that is achieved through trust and collaboration.
Which industries do you think are the best suited to see a positive impact from AI/ML adoption over the next 3-5 years?
AI/ML has the potential to positively impact almost any industry, especially those that are rooted in creativity.
When using a responsible AI tool, those at creative agencies, within marketing departments, and individual designers can turn to AI tools to spark ideas and iterate on existing materials to truly create something unique. AI has the potential to give an initial idea a powerful boost to turn it into something truly special.
However, as AI tools become more prevalent in the creative process, there’s a pressing need to address how this transformation impacts the next generation’s training, skill development, and the integrity of content creation.
Recommended: NVIDIA GTC 2024: Aetina to Unveil Fanless Edge AI Systems Based on NVIDIA Jetson Orin…
Entry-level professionals will learn how to do their jobs using AI and will become increasingly reliant on the technology, whereas generations before learned how to do a skill manually. These individuals will need to be hyper-cognizant of how the tools work and how, if used incorrectly, can do more harm than good.
It is important to note that no AI tool should be left to its own devices.
AI needs to have human intervention and creativity to make sure that it is trained appropriately and certify that the outputs are appropriate for use and authentically represent the real world.
At the end of the day, AI tools are just that – tools – that are only so good as the information used to create them and are not currently agile enough to fluctuate with the dynamic technology landscape. There are changes daily and humans need to make sure that the tools are using the most up-to-date information so that users are protected from potential risks.
What are your predictions for AI/ML and other smart technologies beyond 2024?
AI/ML technology will continue to evolve and we want to make sure that it is doing so in a responsible way. It is important to make sure that people developing and iterating AI/ML and other smart technologies take into account the critical role they have in the dissemination of accurate information. This is especially critical as deepfakes are increasingly popular, creating situations where misinformation is taken as truth and people go down a path that is potentially detrimental to themselves or society.
We will undoubtedly see an increased volume of deepfake content, which will place more burden on audiences to discern what is true vs. fake.
Regulations around AI-generated content, transparency around what is created using AI tools, and the public’s education on how to discern images that are original compared to those that are synthesized will be critical to mitigating the spread of misinformation and decreasing deepfakes overall.
Thank you, Andrea! That was fun and we hope to see you back on AiThority.com soon.
Comments are closed.