Aaron Baughman, Fellow & Master Inventor, Digital, IBM hits the stage at NAB Show New York! Take a look at the key trends and challenges surrounding sports broadcasting and catch Aaron at NAB Show New York this October.

Add AI Takes the Field: The A-List Panel on the Future of Real-Time Sports Broadcasting to your myNAB Show New York Planner.

What are the biggest trends impacting real time sports broadcasting this fall and heading into 2025?

The scope and scale for user demand of content produced within the continuum of personalization will accelerate. On one side, users can control their consumption of content by selecting a particular broadcasting method. For example, they can access content through linear television by selecting a content channel that matches their stylistic preferences. Other methods of real-time broadcasting can be distributed over standards based IT and TCP/IP. This type of non-linear capability that consists of OTT can be enriched by AI or transactional AI. The integration of AI technologies creates derivative content, which provides increasingly personalized experiences.

What challenges or opportunities do broadcast companies need to know because of these trends?

Within the field of Generative AI, which is a subfield of AI, there is both great promise and risk. The power of these large neural network algorithms enable these systems to read 6 orders of magnitude more data within a single month than a human can read within an entire year. With this capability comes the responsibility of balancing each model’s creativity with prescriptive control. Some of the leading areas of open research that broadcast companies should follow include reducing hallucinations and running these models within a trustworthy and human dignity enablement framework.

What’s one thing you wish more producers knew about?

With new Chain of Thought algorithms, model inference times during which incredible self learned skills are created can last between minutes or dozens of hours. This feature is important to understand within the context of live broadcasting businesses so that the correct model and capabilities can be applied to the right use case. Perhaps a smaller generative model that runs on CPU’s is a better fit than a large model that requires GPU’s and several cycles of “think” time.

Further, as story tellers through broadcast, we should keep our entire toolbox open that includes classical machine learning methods. These types of algorithms such as SVM’s, decision trees, and other statistical techniques such as PCA, Eigen vector analysis or operations research capabilities are sometimes the best choices.

What are the top 3 things that attendees should go hunt down on the show floor to expand what they will learn in your session?

1. Explore areas on the floor that combine multimodal techniques of text, video, images, and sound. Potentially seek out examples or concepts where live television will touch all human senses.
2. Discover techniques for balancing technological creativity with prescriptive control that might include LLM’s as a Judge, human moderation, or novel architectures.
3. Find vendors that have real world experience of designing, building, and deploying AI at scale. Consider contrasting and comparing the differences between each approach.

What discussions should they be having with the exhibitors?

As attendees navigate the NAB Show New York floor, have conversations about trustworthy generative and classical AI that are human centric. Both the audience, producers, and anyone supporting the content workflow could leave with a point of view about how to tradeoff AI error rates with real-time delivery and trust. Ask the hard questions about how to apply AI to traditional broadcast television as well as TCP/IP methods such as OTT. Have fun!