Just about a year back, the IET’s President’s address was delivered by Dr. Gopichand Katragadda on the topic “The key to delivering a resilient future”. Needless to say, AI and the recent advancements in generative AI were a recurring theme during the engaging session. At this event, I had the privilege of being a guest speaker and I chose to talk about the AI driven transformation of the Media and Entertainment sector. Predictably, as a passionate AI technologist, I painted a particularly optimistic picture of the potential of generative AI for the media industry. A year since then, with my recent visit to IBC, the premier broadcasting, media and technology show, I’ve been mulling over whether generative AI has delivered on its promise. Bear with me, while I walk through my musings!

There are broadly three areas within the Media and Entertainment ecosystem that can benefit from AI: Content creation, Content distribution and Content Monetization. The early activity in generative AI for M&E focused primary on the content creation space. Given that “generative” AI is generally supposed to “generate” outputs “creatively fashioned” from simpler inputs, the content creation space seemed like a natural fit for generative AI. The somewhat exaggerated predictions of it’s severe impact on content creators stirred sufficient concern to warrant its mention in the demands of the Writer’s Guild of America (WGA) in their strike last year. A year later, there has been some impact on the content creation process but mostly related to efficiency improvement for postproduction workflows. AI used for automated subtitling, translation, dubbing and even lip sync for dubbed content has met with varying success. While the models in these use cases may not work equally well across different content, where they do work, the accuracies can be as high as 95%. For the others, the model inaccuracies necessitate the presence of a human-in-the-loop to correct for errors. It is as yet unclear if AI for these use cases can provide a net improvement in efficiency or reduction in cost. For somewhat simpler use cases that involve creating new formats of content such as highlights, summaries, trailers and other types of short format content, Generative AI models have proven to be far more effective with better adoption rates. 

Among the three areas, content distribution seems to have found the least applications for generative AI. While AI has been successfully used and deployed for encoding optimizations, streaming cost reduction and QoE (Quality of experience) and enhancements, use of generative AI to enhance video quality and information content have not found widespread adoption yet. In contrast, the content monetization space has witnessed far more traction for AI in general and  generative AI specifically. The post-Covid reduction in viewing times and proliferation of platforms has forced broadcasters and streamers to fiercely compete for the same finite amount of viewer time. As a result, platforms are more keenly exploring ways to more effectively acquire, retain and monetize viewers. Not surprisingly, many of these newer monetization models involve the use of advanced AI models for improved user experience, viewer engagement and targeted ad delivery. To name a few examples, multi-modal generative AI for content metadata extraction, personalized content summarization, trailer and tag line generation and intelligent contextual ad insertion. In particular, the use of generative AI models to create personalized and engaging experience for users for sports content seems to have gained traction. Sports content streaming and consumption is one of the largest subsegments within the media industry and represents a large share in both viewership numbers as well as monetization potential. The use of cutting edge technology including AI and AR/VR is transforming how sports content is distributed and consumed. Specifically, there is tremendous untapped opportunity for generative AI in this space with at scale deployments and adoption in likely in the next couple of years.

Looking back at the past year, the rapid technology advancements in video and text generative AI has been nothing short of remarkable. Models today are capable of understanding and generated incredibly nuanced outputs for a wide variety of media content. While many of these advancements have already started to add measurable value in the media and entertainment ecosystem, I for one, eagerly look forward to further transformations with generative AI as the driving force.