- A panel discussion focused on practical use cases around leveraging AI as key to relieving some of the potential fear associated with the technology.
- Moderated by SMPTE President Renard Jenkins, the panel featured Lewis Smithingham, SVP of Innovation & Creative Solutions at Media.Monks; Maria Ingold, Strategy & Innovation CTO at mireality; Quincy Olatunde, VP, Products, Direct-to-Consumer at Peacock; and Samira Bakhtiar, Director of Global Media & Entertainment at Amazon Web Services.
- We will move from an era of monoculture to one which is hyper-personalized with content, brands and other content tailored to our true identity, the panelists agreed.
- Ethically sourced training data and the advantages and risks of working with open source models were also discussed.
A few years ago, when the industry first began to take serious notice of AI, the technology was really an application of machine learning.
The progression and development of the tools and applications since then has been phenomenal, and because of that, says SMPTE President Renard Jenkins, the M&E industry is continuing to find its way through this decade of massive change.
Jenkins was speaking as the moderator of a panel discussion at the IBC trade show in Amsterdam. (The full session, “How to Approach AI and Gain a Competitive Edge,” can be viewed here.)
Amazon has been engaged in machine learning and artificial intelligence for more than two decades, and its practical applicability in the media entertainment space is very important.
Samira Bakhtiar, director of global Media & Entertainment for Amazon Web Services, said: “There’s not going to be one foundational model to rule them all. The ability to leverage APIs to access foundational models, as well as to leverage open source solutions, is going to be paramount.”
She mentioned practical use cases for AI in media and entertainment. One is being able to take archival footage that may not have been shot with a purpose-built slow-motion camera and, by running it through an open source AI model like Llama 2, generate a slo-mo video version of the footage. Being able to do things like super resolution, using an algorithm to up-res archival content to 4K, is another application.
Maria Ingold, Strategy & Innovation CTO at Mireality, suggested that we need to combine subject matter expertise with AI.
“It’s about taking the strength of our creativity and the precision of technology, bringing that together and creating something that reflects what’s happening with society and our value-driven ideas,” she said. “I think what is absolutely essential in order for us to be successful [is to bring] ourselves as subject matter experts to the AI. This is about the human machine collaboration, in order to solve things together.”
Data and Personalization
Quincy Olatunde, VP of products in the Direct-to-Consumer division at Peacock TV/NBC Universal, focused on the importance of gaining an edge in AI by ethically sourcing your data.
“Performance comes down to the quality of the data and how it is sourced. Larger businesses should be aware of the potential risks they may expose themselves to, while ethical responsibility also remains a big factor.”
On this point, Bakhtiar said AWS works hard to ensure that all its customer’s data is secure. “When training a foundational model we create a copy so your models are completely trained with your own data. You can create a virtual private cloud environment around that to ensure its privacy and security as well. We want to ensure that your data is yours. It’s protected, it’s proprietary. It’s your IP; you should be the one that has control over it.”
Lewis Smithingham, SVP of Innovation & Creative Solutions at Media.Monks, chose to emphasize the vast potential of AI to automate and personalize media. He described current the state of AI as “the worst it will ever be,” comparing it to “an affectionate and obedient hamster” with a gambling problem, but also said that AI would kill off monoculture and birth new “microcultures and subcultures that allow people to personalize their content to the greatest scale.”
The level of personalization that AI can churn out, he said, will enable brands and broadcasters “to actually engage with personalization and create completely new channels” based on self-identity, not demographics.
Bakhtiar agreed, saying that if organizations are looking to gain a competitive edge they need to think about how to leverage the petabytes and petabytes of data they have at their command to create those “hyper personalized, hyper localized experiences.”
The Regulation Conundrum
To that end, Smithingham is not a fan of regulating AI, believing that if we keep AI open source it will regulate itself.
“If it’s open source and everyone has access then I think human beings will want something that improves their lives rather than something that ultimately destroys humanity.”
In the end, it’s important to remember that “AI is not an entity. It’s just math, people… It’s vectors and fun ways to make those vectors actually do what you would like them to do,” says Jenkins.
“So remove the fear, get excited about it. And I suggest that artists are a part of the actual innovation and the development of the tool. Be a part of it. Use it, get excited about it. Create something cool.”